It's testing each language's implementation of regexes. A lot of the languages are probably linking in a regex library written in C or C++ anyways. It's evidence that if you're writing something that's heavy on regexes, any slowdown probably isn't due to the language itself but its library implementation.
Indeed, it is the regex implementation that is being tested (and so the language of that implementation is relevant), which is exactly what this discussion is about and also why that comment is careful to be specific about which programs are using which regex implementation.
With PHP at #2? Doesn't seem credible, or the thing being tested isn't meaningfully language-dependent.