I would lean more towards this is one of those truths that can be taken as self-evident. But, at the same time I would be skeptical that any published evidence would account for the infinitude of confounding factors.

My guess is it is similar to IQ results - does a good job weeding out people who know nothing, but does worse differentiating between students who are satisfactory and those who are exceptional

AFAIK there's more evidence supporting that ~B students are the most successful vs A students.