We consider a network implementing scheduling per aggregate flow. We show that the derivation of bounds that was published in [1] is false. We give a worst case bound which is valid for any network topology, given the maximum hop count for every flow and the utilization at every link.