The table that records what is actually analysed.

In this dataset, perhaps we can simply attach an analysis_entities_id to every row by labnr and call it good?

Do not set either of the joined entities to "drop duplicates", or you will get an error message on the join.