Thank you for the example file.
These structures have very weird (if not incorrect) trialinfo structures, which are supposed to label the trials: one column with only NaN and one column with only zeros. The NaN were the cause of the import bug.
I fixed this bug in this commit:
https://github.com/brainstorm-tools/brainstorm3/commit/9393f8fe9df01f70f8efe7b91b421468f23ed01d
Please update Brainstorm and try again.
Make sure you use the process "Import MEG/EEG: Existing epochs".