Hi-
This is interesting. I did a little bit of example testing and had an illuminating discussion with The Bob (with whom *all* discussions are illuminating) about this.
Can I ask what dataset you are using?
3dFWHMx is basically meant to be run on residuals---things that should be pretty deeply detrended already. This is because the predominant usage of this program is to estimate spatial distribution of noise in the data---and residuals are typically the approximation to noise that FMRI produces.
3dFWHMx's detrending is different than 3dDetrend's; as the help notes elsewhere, under the "-detrend .." option:
**N.B.: This is the same detrending as done in 3dDespike;
using 2*q+3 basis functions for q > 0.
3dDespike uses trig functions to remove spikes/detrend (I think it might be something like a finite order approximation to Fourier expansion, with a bit of extra polynomial fitting).
Different detrending will affect 3dFWHMx results, but if you are inputting residuals, the differences should be pretty minimal (hence why I asked what you were inputting, above).
I'm not sure what the calfun error is. If you post the full output (along with input description), that might help.
--pt