because it is not designed as a weed out. there is an unequal distribution of questions, many tests are very different, and they expressly state that it is not intended to be a ranking device, rather it is intended to show that a minimum level of knowledge has been attained.
i have no problem with using scores for ranking, but i do think that the test should be better designed to do that. it would have to involve an even distribution of questions covering each subject on every test, tests of similar level of difficulty, etc...
as it stands now, i think pass/fail should be used. but, it won't happen. some states require a two digit mumerical score, and there is nothing else that is close to giving a comparison between applicants from different schools/systems. so, as imperfect as it is, it is the best we have.