Brad,
I don't exactly understand what you would like to do, but I wonder whether the -glt option in 3dDeconvolve would serve your purpose.
By defining various stimuli and baseline in 3dDeconvolve and setting an appropriate contrast test expression with the -glt option, you could get a 't' map at your will.
For the particular example you mentioned (all stimuli vs baseline), the t value corresponding to each regressor coefficient is already a 't' map. The time points are just implicitly expressed in the intervals spanned by those regressors.
Gang