Agreed demon, a unique constraing on the target table is the only way to prevent this happening.

I've had a look at the data the op put up and I cant see any duplicates on in the set based on the DESC & ID columns.

One thing the OP could do is a Distinct/group by on all columns in a CTE and use that as the Source rather than the raw table.

Personally, I'd throw this back up to whoever generated the file and ask them to explain wht there are duplicate rows, and remove them if possible, or look at altering the extract to make the rows more unique.