You're right, there are no duplicates in the sample, however there ARE duplicates in the original file and thus only duplicates that share the same column 2 and column 3 value and are distinguished between key 1 and key 2 should be removed

The idea is NOT to remove:

1 23456 Scott

2 23456 Stapp

1 56789 Scott

2 56789 Miller

But DO REMOVE:

1 76543 Miller

2 76543 Miller

1 33446 Scott

2 33446 Scott

Does this clarify the question a bit?