Humans are really complex, and something that is proven to work for many on a population scale may not work in every individual case.

And the flip-side is that anecdotes to the contrary do not mean that a population-scale observation is invalidated.

(I know this isn't the point you're making, but others in this thread are.)