It's that, by their own demonstration, the language is nowhere near as good as they say it is. What does "underrepresented" even mean, in this context? It is not used much because it is a PITA to get tolerable performance, where the C code -- and, moreso, the C++ code -- written in the most naive way is near optimal.

I am not going to bother demonstrating the parallel C++ version, but you already know it will be only a little longer than the naive one, and quite a lot faster than the Haskell one.