The Test > Difference of average" works well however and I can see the expected effects.
This is with the exact same list of files selected in Process2, right?
This suggests that the files have matrices of the same dimensions.
This process does not do any checks on the contents of the files, if only averages whatever is in input. While the statistical test processes allow the sub-selection of time windows, signals of frequency bands.
There is a warning that comes up after running the paired ttest saying "File 2 contains a data matrix that has a different size" - it gives this warning for all the files apart from the first one (File 1). This seems odd to me as I am running the paired ttest on the normalised rectified source-level subject averages as described above.
This seems to indicates that there are some issues in the selection of the input data in the process. Your input files may have different time definitions. If you can't open the files simultaneously, this means that you need to fix the time vector of the files first. You can do this with the process: Standardize > Uniform epoch time (ideally, you'd do that at an earlier stage in your analysis).
Or it could be due to another option in the test options. Please post screen captures of the options in the Pipeline editor.