Does it model the observations with 100% fidelity or accuracy? No. Neither does Newton's Law of Gravitation (hence, Relativity), though one would be hard-pressed to call it unscientific.

Newton's law of gravitation can be tested empirically. It can be falsified by experiment. MBTI can't. That's the definition of science and non-science.

Confusion of causality and correlation.

Exactly. MBTI is all based on correlation, not causality. Correlation and statistics can be useful in giving someone the idea for a testable scientific theory. But until you posit a causality and can test it empirically, it's isn't science yet.