Hi, Johannes-
Hmm, it should be possible to do so, yes. Seeing that error surprises me. Would you mind posting the full command you are using?
What might need to be tweaked is that, since the flair has a different tissue contrast than a standard T1w volume, a different cost function might need to be used (e.g., lpc), and maybe a different dataset for the "replacer" face to more closely match the input contrast---though, since the latter gets scaled by the image brightness, the default one might actually be OK.
But please let me know about the command being used, so we can troubleshoot what is happening there.
--pt