Natasha:
I am not expert at the stats involved, but I see at least a deviation from
standard processing that may cause a problem.
You are calling "A" your control task/baseline. That seems to imply
that EVERY (or nearly every) timepoint in your data series is accounted
for by tasks A, B, and C (or A1-2,M1-4 in your other nomenclature), if
you include all time lags, etc.
Unfortunately, as I understand it, the processing stream you are
using assumes that there are OTHER points to be used as
a baseline for comparison. I do not believe that the betas are
meaningful without this baseline.
rich