That's not the point. I don't think anyone is arguing against what he says about worst case for a well-implemented hash table. The problem is the straw-man: all of the 'bad implementations' he has seen, which is all he really attacks.