But the anecdote you're relating doesn't really say anything about the ability of a normal distribution to describe a population. It's just a natural consequence of having a high-dimensional multivariate distribution without accounting for correlation or covariance, along with a low sample size. If you have 4,063 pilots and you take the middle 30% of ten (assumed independent) factors, you're looking at 0.3^10 * 4,063 = 0.02 people who actually fall into that bin, on average -- it's more a dimensions game than anything profound.

That's not to refute the anecdote's takeaway, but the same reasoning doesn't apply to a unidimensional unimodal distribution.