It is strange to see the rm.* files, which suggests
your processing script did not actually complete.
As for the difference, the rm* files should not be
there, so do not worry about them. But look at the
proc script for what the full_mask dataset is. That
is the union of EPI masks across runs, so it is a
slightly dilated EPI mask.
Your commands (3dMean/3dcalc) are trying to create a
full_mask dataset, but you already have that.
See example "c2" from the output of "3dmask_tool -help"
for a nice way to combine masks. Does that seem
reasonable?
- rick