I have no idea what the issue could be. The difference of average and t-test results should indeed look alike.
I would need to have the data in head to test it further…
Could you try to assemble a simple example case and send it to me?
For instance: create a new subject, use the same anatomy as for the other subjects, create two new folders (one per condition), and copy all the files you test (no need to include the 20 subjects if you can reproduce the behavior with only a few).
Then right-click on the subject > File > Export subject, upload the .zip file somewhere and post the link to download it here or send it to me via private message on this forum.
Btw, you could probably optimize your pipeline: saving the TF analysis for all the trials at the source level is probably very slow and use crazy amounts of space on your hard drive. You could run the Morlet decomposition for each subject and each condition separately, and average the output on the fly. This would require some scripting so that you don’t have to run this manually 40 times though.
Thanks
Francois