The logic in https://github.com/sccn/bids-matlab-tools/blob/8128a9d3704a1d20348f62b3fa63608781e66d3d/pop_importbids.m#L520
didn't handle ds003645 correctly. In this simultaneous M/EEG dataset, MEG doesn't have channel locations but EEG channels do. The empty chanloc in MEG caused a hit and caused the code to lookup chanloc from template for EEG channel. And since EEG channels didn't have standard names in this dataset, their chanlocs were cleared out.
The logic in https://github.com/sccn/bids-matlab-tools/blob/8128a9d3704a1d20348f62b3fa63608781e66d3d/pop_importbids.m#L520 didn't handle ds003645 correctly. In this simultaneous M/EEG dataset, MEG doesn't have channel locations but EEG channels do. The empty chanloc in MEG caused a hit and caused the code to lookup chanloc from template for EEG channel. And since EEG channels didn't have standard names in this dataset, their chanlocs were cleared out.
For a different dataset, ds000246, there's no 'eeg' directory nor 'ses-*' directory and the script detected that there's no data because of https://github.com/sccn/bids-matlab-tools/blob/8128a9d3704a1d20348f62b3fa63608781e66d3d/pop_importbids.m#L238
In ds000247, the script failed to import MEG data. Potentially related to Fieldtrip import function