Nothing's wrong here.
First of all, there are sets which simply have identical clones, where e.g. only the internal driver handling differs or just a different name (which corresponds to the IC on the board name).
Secondly, keep in mind what the rebuilder is. It's not set based, it's file based, i.e. it only creates matching files (checksum compare against database) in the destination, not necessary full sets. If you now got for example 2 clones which are 80% identical and the rest 20% aren't, and you only rebuild files which belong to the identical ones, you end up with 2 incomplete clones which are fully identical. The mentioned sets share a lot of files, kovlsjba kovlsjb as clones only differ in 1 file (*prg.rom), maybe you don't have it, and so it rest of these 2 are fully identical.
You're looking at various clones by the way, not at a parent/clone. There is no merge definition between clones, so identical files within clones are pretty common, if they aren't shared with the parent.
And finally, there is a profiler->options setting which allows parsing "merge" information provided by MAME. By default, these settings are enabled (which is good). If you disable them, cmpro does a merge check by not looking at hashes only but also at names...so byte-identical files which don't share the same name are handled as 'different'.
But again, merging is something between a specified parent set and its clones, not clones against clones.