Indeed, this is a known bug in the tool! It occurs when the input science files that are grouped to be processed together have different types (e.g. on-target/off-target offsets to compute the background, or a science-calibratior sequence for VLTI). In essence the type of the first file in the sequence is wrongly attached to all of the others, as well. Work is ongoing to fix it. The file listings, however, are correct. Apologies for any inconvenience that this might cause.