It correlates poorly with any of the competing empirically-supportable models; e.g., as I recall, the I-E axis seems to be a blend of the extraversion and neuroticism axes shared by the EPQ (3 factor) and Big Five (5 factor) models. Better to use one of the scientific models directly than MBTI.
Indeed. But that doesn't mean MBTI is really meaningless, only that there are better models.