This is a general problem with any punitive measure. There will be some false positives (unfair punishments).

I actually do know someone who was not able to get a CS degree because of 'cheating'. His story was that he was working in a group, got verbal clearance from his professor that this kind of collaboration was acceptable, and when that auto-cheating-analysis tool flagged each group members work, the professor claimed not to have said that and was uninterested in fixing the problem.

So, this absolutely does happen. OTOH, I've never personally had any experience with 'borderline' cases. There were enough cut-and-dried cases to keep me busy.

I'm sorry you were unfairly punished.

How should repeated (third or more) plagiarism offenses be addressed if not expulsion?