Well, it turns out that the QBSP fix isn't quite so good as I hoped. Bengt ran it on about 30 test maps with varying results. I do believe that the fix is thoeretically good, but it probably suffers from the same roundoff problems as any other part of the program. I'm going to provide a download for the program on my utils page anyway since it does seem to work better in some cases, just not for every map.