What you should do in the case of the multiple subjects:
- Import all the trials
- Calculate the noise covariance based on all the trials, and estimate the sources
- Calculate the average of the recordings (not the sources) for each subject individually => you will get a source link as well for this average
- Average these source links across subjects
In the files you sent me, I thought the two averages were different, but it was just because of the configuration of my colormap. They are indeed exactly the same.
Francois