You are jumping to conclusions.

Yes, the person doing the import added a small number of (unnamed) peaks at the time. This is weird as we had no such data and the dataset didn't include it (the relevant POI/peak data from swisstopo wasn't open data back then). Some of these peaks were later edited and relocated.

The rest were added much later, without reference to topology or the border, and the simple explanation, applying Occam’s razor, of why they were joined to border is accidental gluing.