Perhaps the BEST B1G Forum anywhere, here at College Football Fan Site, CFB51!!!

Fearless, you're not understanding.  Their backups having higher YPC averages proves my point.  They attained those in far fewer carries.  If you saddle any starting RB with 100 more carries in a season, his YPC average will decrease - it's simple statistics.  Yes, even against worn-down defenses.  The RBs themselves are more worn-down with more carries.  That's why they have backups.
For carries (as with all things) - the greater the sample size grows, the more there is regression to the mean.  

But I'm not sure that's entirely true. There's also usage. Not just how much a player is used, but HOW they are used.
The COP (change of pace) back often has higher ypc than your workhorse back. Does that mean the COP back is better? Or perhaps that he's used in situations and playcalls that have higher chances of breaking for large gains?
Your typical workhorse back is the player that's going to be used on first down in I formation to get your offense ahead of down-and-distance and create easier 2nd- and 3rd-down playcalls. He's the beast that the defense knows is coming but might STILL not be able to stop before he's rumbled for 5 yards. Oh, and that workhorse back probably has a more stout body to stand up to the punishment of all those carries, so he might be your 210 lb bruiser.
Your COP back might get a lot of usage on 3rd downs. He might get a lot of playcalls running in more obvious passing downs. He might get more reverses and sweeps to get him out onto the edge. He probably gets more indirection plays.
You had Felix Jones there who had the highest SEC average. He was the COP back for Peyton Hillis and Darren McFadden. McFadden was a hell of a running back, but the fact that he got double the carries in the same amount of time (2005-2007) as Jones suggests to me that he was being used in a lot of situations that Jones wouldn't have gotten carries in.
Then Vince Young. He had such a high average not because he had low carries, but because he was running when teams had to account for his ability to throw.
Same with Reggie Bush and LenDale White. White was your bruiser that got the ball on first and second down in obvious run situations. Reggie Bush had less than half the carries and a higher ypc by almost 2 yards. I'm not sure Bush would have excelled running into the middle of the line on 1st down.
When you look at the WVU guys, it's the same thing. Cobourne was the workhorse back. Noel Devine was a sprite that was used in the spread-to-run offense and shared a backfield with Pat White who had the same advantage as Vince Young - teams had account for him throwing so they couldn't entirely defend him running.
I don't think it's as simple as more carries = regression to the mean. I think there's a correlation where more carries is a symptom of being used in a COMPLETELY different way than the back who gets a lot fewer carries.
That's one of Melvin Gordon's strengths. He was able to sustain high ypc despite high usage. His senior year he carried 343 times and still maintained a 7.5 ypc average, despite the fact that he was the workhorse and Clement was the COP guy. That's what was so remarkable.

Fearless, you're not understanding.  Their backups having higher YPC averages proves my point.  They attained those in far fewer carries.  If you saddle any starting RB with 100 more carries in a season, his YPC average will decrease - it's simple statistics.  Yes, even against worn-down defenses.  The RBs themselves are more worn-down with more carries.  That's why they have backups.
For carries (as with all things) - the greater the sample size grows, the more there is regression to the mean.  

I understand what you're saying
if I follow, then if Rozier would have had fewer carries, his ypc would have been 8 or 9?
if it's simple statistics then why wouldn't the average, be the average, and bear out for more carries?
If Ron Dayne had fewer carries his average would be higher?  not sure how that works
so if you take Dayne's frosh, soph, junior and senior season, the season with the least carries should have the highest average and the season with the most carries should produce the lowest average?
so if Rozier would have regressed to the mean, please tell me what Rozier's "mean" was

Ron Dayne:
1996: 25 att/game at 6.5 ypc
1997: 20.2 att/game at 5.5 ypc
1998: 24.6 att/game at 5.2 ypc
1999: 28.1 att/game at 6.0 ypc

Mike Rozier: (note: it doesn't tell how many games played)
1981: 151 carries at 6.2 ypc
1982: 242 carries at 7.0 ypc
1983: 275 carries at 7.8 ypc

Cedric Benson:
2001: 18.6 att/game at 4.7 ypc
2002: 23.5 att/game at 4.2 ypc
2003: 21.5 att/game at 5.3 ypc
2004: 27.2 att/game at 5.6 ypc

To put it simply, I'm not exactly seeing a huge correlation here. Dayne's highest average was at ~25 ypc and his lowest average was at ~25 ypc. Rozier seemed to have his average go up with total number of carries, but I can't tell how many games played from Wikipedia. Cedric Benson's worst year for ypc average was his second-highest att/game total, but his highest average was his highest att/game total.

I'm just finding it really difficult to find a correlation here. 

I personally find my explanation--workhorse backs get more carries, but they're a different usage style--a lot more credible than the idea that more carries in itself is "statistically" going to force the ypc average down. 

Let's take Rozier, since his YPC only went up with addition carries from year to year.  

1983 - Rozier has the most carries and a 7.8 ypc.  Gill, the QB, is excluded from this, because that's a different position and QB's rushing averages tend to be lower.  But UNL had Jeff Smith run the ball and Mark Schellen.  One of those might be a FB, but whatever.

Rozier, at 7.8 ypc, should've gotten the ball more.  The next 2 RBs averaged 4.9 and 5.6 ypc.  So to maximize efficiency, Rozier should've had some of their carries until their YPC averages infalted to around 7.8.

Now we assume players improve over time, at least they're supposed to, but we also know good OL graduate and all that.  With a big-enough sample size, that's all a wash.  But we're looking at one team here, so I can't control for those right now.

1982 - Rozier has the most carries, with a 7.0 ypc.  Backups are Craig at 4.9 and Wilkening? at 5.2.  Smith is there, with 56 carries for a 10.2 ypc.  I'm certain he only had garbage time carries, and that is a thing, but with that ypc, he should've gotten more carries.  Rozier should have as well.  Smith, because giving him the ball until his YPC decreased to a number closer to the others would be maximizing efficiency.  Rozier because he had the best YPC of the top 3 ball-carriers.

If this is all confusing, I'll try to make some simple, logical statements.
1 - Mike Rozier was Nebraska's best RB in 1982 and 1983. 
2 - His YPC alone does not make statement 1 true, but his YPC and his volume (number of carries) does.
3 - With fewer carries, his prominent backups would very likely have had higher YPC averages.
4 - Jeff Smith should have had more carries in '82 and fewer in '83.  

The point is, Rozier could've handled 50 more carries, 100 more carries, and his YPC average would've remained higher than his teammates'.  That's efficiency.

Taking from the lists, and noting the leader at every 100 carry benchmark:
1000+ Dayne, WIS (5.8)
900+  Griffin, OSU (6.0)
800+  Williams, TEX (6.3)
700+  L.James ORE (6.6)
600+  Gordon WIS (7.8)
500+  Gordon WIS (7.8)
400+  Love STAN ties Gordon, but Gordon has more carries.  We'll see what he does this coming season.
300+  Weatherspoon (HOU) 8.2

*Gordon is the only repeat name on the list, and with only one guy tying him, no one surpasses his YPC for 300 carries, which is nuts.

*Also of note, look at how the YPC increases when you go down 100 carries - 0.2, 0.3, 0.3, 1.2, 0.4.  Gordon's increase is 4x the average of the other benchmarks.  There is a very good statistical case for Gordon as the best ever.

This is what makes Gordon such an outlier (in a good way) - no RBs with a higher YPC at 500+ or 400+ carries.  There "should" be, but he was so good, there isn't.  Yet, without fail, we find a higher YPC average down at 300+ carries.  
And I'll make you a promise - if we looked below 300+ carries, we'd find someone with a YPC over 8.2......because math.

And with Rozier....there would be a point in which you'd give him too many carries and his YPC would start to fall.  I don't know how many it would take, but it's obviously more than 275.  Say you gave him 400 carries in a season - is anyone going to suggest his YPC wouldn't drop?  

I could make that suggestion based on 2nd halves vs poor teams such as Kansas or others were he averaged over 9 yards per carry but was pulled from the game early because of the huge lead and the risk of injury.  If he simply doubles his carries vs weak opponents his average goes up as the number of carries goes up.
for example, 1983 vs Kansas - in the 1st half Rozier had 26 carries for 230 yards 8.8 average, in the 2nd half he added 5 carries for 55 yards - ave 11
I'd bet $$$ the 5 carries and 55 was the first drive out of the locker room.  I'd also bet that if Rozier would have had 52 carries total he would have averaged over 9 yards per attempt.  But, we'll never know.
there's no possible way of telling what might have happened based on the stats of any of these great backs.
there is also, no possible way of predicting if Melvin Gordon would have averaged more or less than Dayne had they each had the same number of carries.  I'd bet on Melvin having the higher average at 1,000 carries, but there's no way to tell

We have no way to tell among these individuals.  But the more data you collect, the clearer the picture is - the higher the volume, at some unknown tipping point, you start to regress to the mean.

the problem is that we are talking about individuals
and as you said, there's absolutely no way to know, regardless of stats
with any guy on the list, his next carry could be a 3 yard loss or a 73 yard gain
way too many variables, such as play calling, quality of opposing defense, field position