The missing scaling info was:
[* 0.00100329] -26.548 to 30.0715 [scaled]
I guess I was under the impression that after using 3dmerge all of the voxels below the threshold criteria would be 0, even after the Talairach transformation. This is how it appears when viewing the Patterns_act1tlrc data set (fim data set, Func listed as 0 for the voxels in question), but there seems to be values greater than 0 when 3dmaskave or 3dcalc is used. Normally I threshold data (3dmerge) in the original format and then apply the Talairach transformation. Would it be more accurate to transform the functional data set first (results from 3dDeconvolve) and then apply the threshold (3dmerge)?