Because KF doesn't go for linking to blogs, if you want to find the research behind this, you're going to have to use Google. However, the idea that increased minutes = decreased production per minute was debunked last year, when people started questioning the value of PER due to its dependence on per-minute stats. Three or four different studies of the subject, looking at large samplings of players, had findings like: increased minutes will result in increased production per minute 70% of the time, decreased production per minute 28% of the time, and level production per minute 2% of the time.
So it depends on the player and the circumstances, but about 70% of the time, more minutes will = higher per minute stats, not the other way around.