It would be interesting to see the benchmarks run again once that bug has been dealt with - to see by what amount fixing the issue causes the framerate...