The latter. The tools in question (Tamarin and ProVerif) use a model of signatures that goes back to 2000/2001 which is around when these key substitution properties were discovered. Consequently, they were missed from the models and because the properties themselves aren't that well known, it took a while before it was noticed.