The problem isn't really "the benchmark is flawed", the problem is that most of us seem to wish for data that tells us more than it can, magical wishful thinking on our part - Will the application I haven't designed yet be faster, and scale better, and be completed by us faster, if we write it in X or Y? :-)
