Probabilistic cloning with supplementary information

We consider probabilistic cloning of a state chosen from a mutually nonorthogonal set of pure states, with the help of a party holding supplementary information in the form of pure states. When the number of states is 2, we show that the best efficiency of producing $m$ copies is always achieved by a two-step protocol in which the helping party first attempts to produce $m-1$ copies from the supplementary state, and if it fails, then the original state is used to produce $m$ copies. On the other hand, when the number of states exceeds two, the best efficiency is not always achieved by such a protocol. We give examples in which the best efficiency is not achieved even if we allow any amount of one-way classical communication from the helping party.