you're arguing a tangent to GP's point, which could be rephrased as "assuming we're already willing to spend $x on welfare, we can structure the payments in a way that effective income increases monotonically as the recipients earn higher wages for work". you can fix welfare cliffs without spending any more/less money.