String manipulation is the core functionality for all languages so this allows to compare languages fairly.
If that doesn't clue you in on how utterly full of shit he is, I'm not sure what will. I mean, Java Strings are immutable. This test is just about adding strings together. That's pretty much the worst possible case for trying to benchmark Java. So if you're coding Java and adding a bunch of strings together, you use a StringBuilder and not a String. Only you can go look at the source code, and whoever wrote it didn't. Not only that, but how much memory Java would use during the run would depend pretty much entirely on flags given to the JVM, because it would just keep eating up space copying the immutable String over and over until it was forced to garbage collect. And that's all just a quick inspection of the Java comparison. I am pretty confident without looking that that margin of difference between C and C++ is entirely due to pathological C++ code.
I mean really, if you think that your interpreted language is comparable to any major compiled language in performance, you're an idiot. Sabotaged test results (whether the result of duplicity or incompetence) don't change that.