At least in C64 Duncan made some sort of tool that will open the roms (in various formats) and compare them file by file (i think!).
Also, 99% is not that good
It all depends on the total size, system and so on. You can't easily or clearly know they are all the same software, wrongly dumped or hacked to create more dumps, versus original different versions. On the other side it might be a good way to differentiate between software titles based on that similarity.
Sets with a similarity of 99+% are probably the same software title, still they can be a different version, modification or just bad dumps.
In a full disc of 700MB, 1% different is 7MB which could just be the executable file (an update, cracked version or something), while all data was the same. On the other hand, older and smaller sets didn't had much more data so a difference of 1% could just represent a few changes in a savegame or executable, right?