Q1: It does not make sense to use amplitude modulation with IM, since computing the coefficient (beta) for the AM component requires at least TWO events -- one for the beta describing the mean response, and one for the beta describing the variable response. With IM, there is only one event per regressor. Instead of AM, you get a different beta for each event time in the same stimulus file -- and then have to combine those betas somehow to get useful information.

Q2: Yes, only one IM regressor is allowed in 3dLSS. This program was a quick implementation for someone (and now I don't even remember why I wrote this code). I am unlikely to make any changes to it, since 3dLSS is very low priority for me.