On the axiomatic treatment of the phi-mean