Not significantly, no - but comparing the two, RVO2 does seem to behave a bit more sensibly under stress. Your algorithm seems to make crowds noticeably swirl in a clockwise direction in very crowded areas. And it feels like agents collide slightly less often with RVO2, making movement look a bit smoother (that's just subjective of course - crowds in reality don't look as smooth!).
There's no rush - I've got an implementation here that's (almost) working, I just thought it would be interesting to see your take on an integration! (interpolation is such a pain!)