I really don't understand the questions you are asking me, as that don't appear to be related to the statements I made, but I can try to respond to your general comments.
No matter how the training is set-up, it will not be realistic. That's simply recognizing that this is a simulation. Whether one approach or another is more "realistic" is largely a subjective evaluation of each user with regard to what they consider to both be realistic and important. While I fully agree that the current training system in BuzzerBeater is not "realistic", I don't find what you have described to actually be very "realistic" either. But a debate about the subjective nature of "realism" is pointless, although that largely appears to be what is behind your suggestion.
Your proposal makes both training and game planning easier simply because it disassociates them from each other. Doesn't really matter what other options you try to add back on on either side to increase the difficulty of each one independently because you have still completely removed one of the biggest strategic elements from the game by separating training from the games. Having training linked both to the minutes a player plays in a game and the positions they play at is one of the most important strategic elements to this game, since to really train a player well, a person has to play him out of his best positions at times, which will reduce the strength of his team on the court. For example, making those decisions around which games to train a center at point guard to increase is passing or outside defense is one of the key elements of the game. Your proposal removes that and replaces it with a number of proposals which I think involve little strategic thought or planning. Therefore, in my opinion, they lessen the game. Whether or not they are "realistic" is very much a tertiary consideration.