Once they are in college, the impact of parents is quite minified. So if you have a stupid whose default academic potential is 10 but has a parental multiplier of 2 for having involved parents and another kid whose default academic potential is 15 but has a parental multiplier of 1 for parents who don't care, once they get into college and the parental multiplier goes away the one who has a 15 should do better than the one who has a 10.

Now, numbers aren't that easily available. And maybe parents caring has a permanent effect that lasts even once they are gone. Real life is far more complicated than my example. But it shows a very simple possibility of why measuring potential while ignoring environmental failures can lead to a sub-optimal ranking.