Difference in brain area/size of tstat from FSL randomise output


I have a question regarding the masking using FSL randomise with TFCE. I have noticed that the brain area of the tstat map output from randomise differs when it is run from the terminal vs. the GUI (even when running it from the terminal does not replicate the overall area).

When running from the terminal, I usually use a bash script, where I first concatenate the subj-level cope.nii.gz files and mask.nii.gz with fslmerge; then, running randomise. I tried running randomise twice from the terminal (each with slightly different commands to trouble shoot a separate problem) and have noticed that the tstat is larger (in terms of brain area - not the tstat values) in one case.

Commands used for the respective randomise runs from the terminal:
randomise -i filtered_func_data_cope1.nii.gz -o /output -T -1 -n 5000

randomise -i filtered_func_data_cope1.nii.gz -o /output -m mask.nii.gz -d design.mat -t design.con -e design.grp -n 500 -T --film

I also tried running randomise from the GUI and again noticed the difference in the brain area of the tstat (again, not the tstat values).

Command from the log file:
/usr/local/fsl/bin/randomise -i filtered_func_data -o stats/ -m mask -d design.mat -t design.con -e design.grp -n 500 -T --film

That has led me to think that there is something odd going on with the masking (?).

I have attached some screenshot - although, I am not sure how well you can see the difference.

Does anyone have insight into why this is the case?