One of the main advantages of CRC is that the order of the bytes does indeed matter (as opposed to simple sum/xor checksums).

If you actually get a match this way, it's coincidence (collision) or your input data is structured in a way that it doesn't matter in which order it is accessed (all bytes have the same value).

However, if a CRC algorithm accesses the bytes in random order, it's either trying to do something "clever", or it's simply buggy.