This appears to be correct after all, so nevermind. The NY Tax tables have some non-linear jumps when your AGI crosses certain boundaries. And that appears to be what is happening here.
