The program sometimes has trouble with datasets whose values are spread out over a large range of numbers, as your datasets apparently are. One possibility would be to compress the range. This could be done with a simple 3dcalc command, as in
3dcalc -a rot23dbM+orig -expr '.01*a' -prefix rot23dbM_c -nscale
Scaling down by a factor of 100 (since your upper limit is about 32000) will make the histogram simpler, and that may help the program think. The program was originally written for datasets whose numerical ranges is a few hundreds, not tens of thousands. I suppose it should be patched -- but give this a try.
bob cox