What's wrong is that they measured the performance of Lisp programmers vs Java/C++ programmers. This is fine but you're not comparing the merits of the programming language. The Lisp programmers would win even if they were forced to program in Java.