Craig,
I'm very sorry that I misunderstood the situation of your experiment design: I thought that you purposefully made one column of zeros in the stimulus matrix so that you would not analyze it.
Currently 3dDeconvolve gives a warning with a stimulus of all zeros, but it would stop if the user wants any glt tests. Doug Ward probably had some reason doing it this way. We will take a closer look of the details and evaluate the situation before we make any decision on whether it can be improved.
Thanks,
Gang