I don't know much about how these scanners work, but I assume that they look for a sequence of bytes (code) that they recognise.
If the gets reports of false positives, I imagine that they have a CRC whitelist which they simply add to when you report a false positive.
They cannot change the 'byte sequence' because they would have to go back and check it works on all previous (real) viruses still.
No, that's not the problem here. They have signatures of known viruses in the form of known byte sequences or similar that their scanners detect and such detections work pretty safe, there is no such detection triggered here. The problem here is the "heuristic analysis". They usually have a "point system" which could be triggered by almost anything these days. They don't recognize known byte sequences, or at least not as the only detection pattern, but rather combinations of lots of patterns like for instance API functions that are typical for malware to call, embedded data of certain sizes, patterns that look like executable code within embedded data sections, certain words and texts, small changes from known-good packages (to detect repackaged software) and similar. When combinations reach a certain total level, a "suspicious" detection is triggered in the scanner.
You can tell such detection by malware names that contain "susp", "gen" or similar. Such detections are not about any known malware, but simply something that could possibly contain malicious code.
From discussions with antimalware vendors I have learned that it's important to file false positive reports when this happens. That's the only way they could tune their detection logic to be more precise and so that they could avoid false detections with future versions of similar software. If, on the other hand, a known "named" virus or similar is detected it could be possible to avoid the detection by rewriting parts of the software to avoid a particular sequence of bytes to appear in a place where it's detected as this particular virus. But this happens extremely rarely I would say. It has never happened to me and I don't very often read about such cases in forums either.