If you agree that there are hidden mechanisms that shape the inequality, then claiming that “only considering the factors that effectively prevent studying” is fair would require a great deal of certainty in knowing these factors. I don't know how effective “positive-discrimination” is in practice but to me it seems that its principle is to do what you say but with discriminatory features as a proxy heuristic for these “effective features” (this makes sense if you believe they are sufficiently immesurable IRL).