Hi Tim,
Yes, I think that you would get better activation maps if the experiments did include some resting periods for the other conditions to compare with.
With the current experimental design and with all nine conditions covering all timepoints, there is no way to model the resting state directly, and you would have to sacrifice one condition (i.e., Cond3) and make it as a baseline condition since 3dDeonvolve requires such such windows for modeling the baseline. Then all other eight conditions would compare with this baseline (Cond3), and you would also lose the opportunity to test the significances of this condition (Cond3).
Is this approach acceptable to you?
Gang