When you upgrade versions, especially when you go from LTS to LTS, you get changes in the rules. Rules are added, deprecated (sometimes removed, but rarely), and very often existing rules get smarter.
Specifically, that means that false positives are eliminated and so are false negatives. So more than likely that’s what happened. Generally such new issues are backdated when they’re raised on the first analysis after an upgrade. However, if the issue has secondary locations (locations that lead to or contribute to the issue) it’s going to get the date from the newest 2ndary location, which could put it in the new code period.
Could you take a look at some of these new/old issues and see if that makes sense in their context(s)?