Runtime semantics of p:try

An XProc user pointed out to me that the following sentence (Section 2.1, Steps) is not entirely correct for p:try:

"The runtime semantics of a multi-container step are that it behaves as if it evaluated exactly one of its subpipelines. In this sense, they function like compound steps."

Clearly, p:try may evaluate (parts of) both sub-pipelines.

Regards,
Vojtech