Yes, if you pick 4 attributes (almost any attributes) then they'll correlate to something.

Myers-Briggs tends to emphasise classification, which implies bimodal distributions. In reality, the traits it measures are more normally distributed.

Also, every attribute but introversion seems pretty meaningless. If it were a scientific model, the traits would be open to debate. But people treat it like an astrology chart that's been handed down through the ages. It's not the initial attempt at creating a framework which I'm critical of, but the way it's so uncritically accepted.