A couple of questions:
(1) I see that you explicitly model rest as a regressor. Do you have baseline blocks in your experiment in addition to rest?
(2) Why not model before treatment and after treatment as two separate regressors?
(3) Since the measure seems at trial type level (not individual trial level), it seems to me you should NOT model the measure in individual subject analysis with -stim_times_AM2. Instead I'd use the measure as a covariate at group analysis level using 3dMEMA or 3dRegAna.
Gang