You can define a limit, at what duty cycle does the cost savings reverse, that would be a solid counter argument followup I would think. e.g. at 0.1% bang-bang, how does it compare when you look at similar energy consumption vs the equivalent PWM, If you really wanted to stick the point you could even graph the net gain in airflow vs power consumption with duty cycle control vs bang-bang, as that spin up energy is lost every time in bang-bang