Incremental Refresh Error

I'm experiencing this exact same issue.  Incremental refresh is configured properly (query folding, etc).  RangeStart and RangeEnd filtering on my fact table using a datetime column.

After reading the documentation (several times), my understanding was that it would simply overwrite (delete/insert new) any and all rows that qualify on the datetime column filtering.  If this were true, then I would not have any duplicate rows in my data model.  I have confirmed in SSMS that my actual table only contains one row for a given primary key.  Yet the PBI incremental refresh engine is telling me there are duplicates.

The documentation mentions nothing about honoring any kind of primary key relationships.  The documentation mentions nothing about intelligently identifying rows that have UPDATEd, as distinct from rows that are new and should be INSERTed.

WHY?????