The correct behavior of a service composition depends on the appropriate coordination of its services. According to the idea of channelbased coordination, services exchange messages though channels without any knowledge about each other. The Reo coordination language aims at building connectors out of basic channels to implement arbitrarily complex interaction protocols. The activity within a Reo connector consists of two types of communication, each of which incurs a delay: internal coordination and data transfer. Semantic models have been proposed for Reo that articulate data transfer delays, but none of them explicitly considers coordination delays. More importantly, these models implicitly assume that (1) internal coordination and data transfer activities take place in two separate phases, and (2) data transfer delays do not affect the coordination phase. This assumptions prevent maximal concurrency in data exchange and distort the evaluation of end-to-end delays in service compos...