@sgmunson - Sorry for the bad test data and ambiguities.

@ScottPletcher - I've tested your approach and it's much faster. It runs 5 seconds for my test table with 100 entries....