The new normalize option is
not ready. At this moment, I'm working on incorporating waver and basis expansion deconvolution into 3dDeconvolve, so that you specify the stimulus timings directly with an option like
-stim_times 1 filename.1D GAM
where filename.1D would contain the times of the stimuli and GAM specifies the response model (there will be several models available, including the one that SPM uses). This work will probably take the next few days -- fitting it into the existing program is a little tricky.