Might be 70%. I think you might be missing many many things you cannot simulate in practice.
I generally subscribe to the idea that skills and physical attributes are mostly developed/improved in practice and the offseason. However, the efficacy of practice and the offseason in developing skills is a function of playing time and in-game performance in at least the following two ways:
1. In-game performance feedback is a better measure of current effective (in-game) ability/skills than practice performance. This feedback is essential in tailoring a player's training regime/focus, and should have an effect on the speed, direction, and overall efficacy of the training.
2. Players are people too, and their motivation and overall engagement are affected by the opportunities and responsibilities they're given. A player who feels valued and sees his hard work paying off might be more motivated to put in more time training -- and make that time more effective. He might also be more engaged when he's playing, as in Lyles' case (yes, there are other potential explanations).
Now, it may be that the first effect only affects the speed at which skills will be inevitably developed, and so it's not terribly important -- although CBA realities need to be considered. Further, it may be that the group of marginal motivated/unmotivated players is small -- that is, that the vast majority of players either are or aren't engaged/motivated, and as such not affected by their lack of playing time. That's a (tricky) empirical question I can't answer at present. However, based on my experience teaching, in the mission field and in other work/school environments -- admittedly, surrounded by regular/mediocre people rather than professional athletes -- I find that hard to believe.