Personality traits typically predict behaviour across situations at around r = .30, which leaves most of the variance unexplained. Why is that modest figure taken as support for the trait approach rather than a refutation of it?

D. Because traits predict aggregated behaviour far better.

The .30 figure comes from predicting single acts, and even a very sociable person stays quiet at a funeral. Epstein made it precise: averaged across twenty social occasions the correlation is two or three times higher, with nothing about the trait changed. It is why we know our friends' personalities well and predict any given evening poorly.