Moreover, for individuals scoring in high percentiles, to which you probably belong if you find the stuff written on this blog interesting, there are strong diminishing returns to high scores even statistically.

This doesn't seem to be so up to at least the 1 in 10,000 level. However, I agree that the predictive power of theses tests is still small relative to the remaining sources of variation (although it is one which we are relatively good at measuring) and they shouldn't be over-weighted.