Hello,
you can certainly model your network using a single queue per node, which models the sequence of messages sent to that node, but not yet handled by it (its "inbox"). Senders append their messages to that single queue. This may make your specification a little easier to read. However:
1. This representation enforces accidental precedence between messages. If both A and B send a message to C then in the model with a single queue, these two messages will be ordered, whereas in the model with N*N queues they are not. Whether this matters or not depends on the system you are trying to model.
2. The single-queue representation doesn't help reduce the number of states that TLC has to compute: it's not clear to me what you mean by "reducing[ing] computing resources". (The memory footprint of a single state may be a little smaller, but I doubt that you get a noticeable effect from this.)
–––
Rereading my previous message, I see that there was a mistake: as you probably found out, the correct definition should have been