I don't think it is really known what the factor is.
If you did get proportional training, then you could maximize your team's training by not having any players get more than 48 minutes, since any time over 48 minutes is wasted in terms of training. On the other hand, there is a risk in not letting players play 48 minutes, such as losing games and ticket revenues, and a negative effect on game shape and stamina.
To me it would seem silly to make a discontinuity at 48 minutes (eg someone who plays 47 minutes gets 3/4 as much training). But there are other curves that avoid a discontinuity, but provide some penalty for playing less than 48 minutes. For example the square of (PT/48) would give 96% for 47 minutes, 56% for 36 minutes, 25% for 24 minutes.
If I get two players at a position close to 48 minutes in the two league games, I tend to rest them in the scrimmage. If for some reason one is around 35, I'll use him as a backup in the scrimmage - but that is in part because I feel it is more important to make sure my better players get maximum training, rather than simply getting maximum training for the team as a whole.