I haven't read the original paper, but presumably they do a regression, which can pick out trends despite a few exceptions. Taller people tend to be heavier than shorter people; the fact that there are some tall, skinny people and some short, fat people doesn't falsify the general height/weight relationship.

As for discrimination, cultural discrimination against women suggests that discrimination against women may be a factor in the test scores. Reverse discrimination is a possibility, but that seems less plausible given the larger cultural context.