You can train a skill and improve on it. You can't train your head shape or arm length nearly as much. The brain is more malleable than some of the other body parts.
I'm not sure what you're suggesting then - we should train all pilots to be uniformly brilliant at everything so there's no variation and we can use any pilot to test, before we test any more aircraft?
Right, but my point is there may not actually be any pilots who are merely competent at all skills. The distribution of skills may be more complex than that. You may be making things safe for someone who does not exist, which is pointless, and you'd be better off making things safe for something other than a simple mean average.
You're going off on a tangent based on semantics and speculation. Pilots who are competent at every basic skill, but don't have years of combat or test pilot experience or extensive experience on the plane in question absolutely exist and that's what we're talking about.
You're getting hung up on this, like looking for the individual in a population who exactly matches an average, while ignoring the normal distribution that surrounds it. Don't obsess over the point, just center your window function around it.