Bad Theories that Track Robust Regularities

This interesting (as always [that is, I want to have millions of Malcolm Gladwell's babies]) Malcolm Gladwell essay on personality tests comes down hard on the Myers-Briggs. Now, I think he's right about everything he's saying about personality. Yet it remains that the Myers-Briggs does tend to track some fairly deep and important regularities. For instance, almost everyone I know is an NT. (Indeed, almost all libertarians are NTs, which is helpful for understanding why we do such a terrible job communicating to non-NTs.) So Gladwell's right, but I sort of believe Myers-Briggs anyway. What's going on?

Similarly, I think the mind is pretty modular, if not massively so. Thus, I don't think there is anything like a general reasoning capacity. There are various cognitive subroutines that are elicited by different environmental cues, and it is possible to reason effectively in one domain while it is possible to make systematic errors in other domains. So, in terms of foundational theory, I shouldn't believe in G, general intelligence. Yet I'm rather impressed by the data on G, and how it predicts quite a number of things. So what is G tracking if there is no general reasoning capacity? And what is Myers-Briggs capturing if personality is more continuous and labile that the test seems to assume?

(Bonus conjecture: If you have taken the M-B test, put your personality type in the comments. I bet that over 70% of The Fly Bottle readers are NTs of some sort. I am, FYI, an xNTP.)