I don’t think that a clearly mismatching DNA has anywhere above a one in a billion chance of belonging to the same person. So either the sample is wrong, which

I don’t think that a clearly mismatching DNA has anywhere above a one in a billion chance of belonging to the same person. So either the sample is wrong, which is near impossible with a rape case nowadays, or some technical mishandling (switch of samples?) that can of course happen but this is not likely leading to a 5% error rate. Maybe in the early days people used erratic methods for reading pieces of the DNA but it’s hard to imagine a 5% error rate even then. I guess the other way around could work, ie a match may be a twin sibling or something.