Okay, I found the Pubmed citation, and it’s not definitive either - indeed, the title doesn’t accurately describe findings detailed in the abstract.

We seem to have run down a bit of a blind alley on this, but I think we can agree that it’s very risky at best. Interestingly the primary finding of the studies was that HIV can be transmitted even using coitus interruptus properly, and indeed higher than the risk of pregnancy.