Ready.
As copycat82 is an ignorant plagiarist, so often we may understand what some erroneou statement was supposed to be stating, only after we find the point of correlation in the copied prior art source.
The case of the absurdness of the components "starting immediately" (that is, zero wait) but also planning "time Petri nets" seem oxymoron. But then again, if we notice that Da80 is the source, we understand that.
Obviously, copycat82 was trying to talk about imitating the structure of the two-level sifting, as found in the X-transition of Danthine (modifying the X-transition of Nutt, with an extra rule) -- specifically, if there is no token in B, the token in A is send to abyss, not used. After that first test/sifting, is the time Petri net (that may wait).
But the plagiarist copycat82 has messed that up. In the Danthine case, that X-transition is (redefined after Nutt X-transition as) a primitive of the net formalism, and the token is waiting at the entrance, until the time Petri net may or may not fire. But the copycat82 "component" is macro, that is, the first step "immediately" gets the token to its internal. Thus, as far as that abstraction is concerned, the token has started activity. Not waiting.
But internally, the ordinary Petri net transitions wait for firing, until the token is output from that transition (and finally, from the macro). Thus, copycat82 was supposed to carry the time-Petri-net waiting, to the first internal transition (or, distribute to a few of such, internally). If that were stated such, that was obviously imitative of Danthine, trying to trivially emulate Danthine'ist two-phase firing-testing, with macros. That is, again, I have found that the cause of a vagueness/error of copycat82, is pointing to its plagiarism -- and again, most directly the formal structure of Da80.