Ready.
On the page 38 of copycat82, there is only the "Fig.3.2. An example of a modified Petri net." This corresponds to the "Fig.2" of copycat83 (on its page 736).
On both sides, that is, both C2 and C4, precede C3, which precedes C5 and C6. There is no further need to draw extra arcs, depositing tokens to useless places, but copycat82 does that, when there is an arc out of C2, depositing a token to a place before C5. That is a useless act. It does not enable anything. Symmetrically, on the other side, the C4 and C6 also have such a useless multiplicity of paths between them.
It is a loop. Yes. But C2 and C4 write-only, and C5 and C6 read-only. i.e: No feedback. This does entail that, unless totally at random, C2 and C4, must preserve internal state.
In fact, preserving-internal-state is a fine thing about E-nets, as Da80 discusses, too. The problem is with those data-item boxes, which present possibly false commentary. The arrows point such that, the feedback possibility is excluded. An argument that "It makes an error obvious." is itself obviously faulty, because that commentary, has not saved copycat82, and also copycat83, as the "Fig.3.2." is published in both of them. This could be an unimportant example, but the rest of the copycat82 is full of such obvious-faults, too. It appears, a Ph.D. team of three or four people, not to mention the Ph.D. jury, and paper reviewers, could not suffice to correct these. cf. The resignation of Mr.Parnas, and his articles, in early 1980s, because he did not believe computer-systems were dependable enough. The examples I present about copycat82, may relate to what he had said.
If that is a "message-based" system, what does it mean that the same data-item is read by two separate subnets/"component"s? A first guess is a "similarity" representation, but such vagueness, throughout its graphs, fail any explanation. The page, Similarity semantics of data? at this site, discusses it.
This example uses only "and-input" and "and-out" (which are basic Petri net behavior) except two "xor-output"s (which corresponds to the X-transition of E-nets). As such, it is just like Danthine (1980) in extending Petri nets with the X-transition of E-nets. This is a remarkable point, because all other types of examples are false, in fact. In other words, copycat82 is faulty, whenever its examples have anything other than the ordinary Petri net transition in/out behavior (join-from-all-inputs, fork-to-all-outputs), or the X-transition ("xor-output"). As we discuss the case here, such examples still have their problems, but that is about other, more subtle interpretations, where it is painted, with data boxes. i.e: Again, where it is ever so slightly different than Da80, it is faulty.