I completely agree with Martin Fowler on this one. Better programmers out perform worse once. (And that is why they are cheaper in the end.)
It’s a very nice example of More is less.
Maybe one slightly change to his article: in an truely agile team, when the better programmers pair with the lesser ones, they will bring the lesser ones up to speed. So you might do ok with a few lesser programmers in your team. I said a few, that means less then 50%. More like 10 to 30%. (Otherwise the quality of the team is too low.) When these lesser programmers are brought up to the same speed you can hire other lesser ones. It is my guess you won’t need them anymore.
I would chooses 10 smart people over 100 dumb, every time. I know the smart team will outperform the other team big time.