Counts and measures wrong after data refresh

So, the problem must be that the addition of new data into the model has somehow affected existing relationships so that the data in a table is somehow automatically being filtered in measures and visualizations. And that table is then automatically filtering the same in another table. But I cannot understand the logic…

The tables that are being filtered are circled in black below. The table circled in red is the node that connects the data in this section to other data sets in the model.  If I disable the relationship indicated by arrow, then the calculations seem to be correct again. But if I enable the relationship again (which had no effect before), then the calculations are incorrect. This relationship is not new!

Below you can see the new data that has been imported via new queries, drawn around in red. The same node that is the connection between this data and the previous is also circled in red. And the points of connection between that table and the new data are circled in black. The relationship indicated with a red arrow used to filter in both directions. I have changed it to one direction. But that has had no effect. Note that with that change, the new data does not filter the original data in the model. The old data filters the new data. Given the direction of the relationships, I do not understand how the new data is affecting the old data…

If I disable the relationship indicated with a black arrow, so the other side of the same node, then the measures and visuals are correct again. So, the problem is definitely the relationships with the new data. (This relationship is also not new. Only the circled tables and their respecitive relationships are new.)

The model is quite complex. (There are several other sections of data, all of which are connected through the same node (which is a reference table that contains the unique identifier for each person.) So, I could be missing something really obvious…