Hello Guiomar, we just implemented an automatic way to generate the required montage to recover the reference channel.
After that, you still need to apply the montage, to create a copy of the EEG data with the reference channel added. As expected in this new copy, the reference channel is flat in 0 V, but if we use average reference we retrieve its time series.
Before computing the head model and estimate sources, the location of the new reference channel needs to be added.