Then you should be able to make the label refer directly to that physical encoding of M. ... you ought to be able to refer just to some specific physical system T that "encodes" M ... if you're still unhappy with my label, then you would probably be unhappy with this unpacking of your reference to M. But I can think of no other way to make good your claim to refer only to physical things.

Well, I would need to permit more than just one physical encoding; I'd need to permit any physical encoding that is, er, isomorphic to an arbitrary one of them. But I don't see this as being a problem -- it's like what they do with NP-completeness. You can select one (arbitrary) problem as being NP-complete, and then define NP-completeness as "that problem, plus any one convertible to it".

So it appears I can avoid binding the meaning to one specific physical system, while still using only physical referents. And yes, your updated terminology is fine as long as you allow "symbols" and "fed" to have sufficiently broad meanings.

Incidentally, are you saying the same problem arises for defining "waves"? Do you think that referring to one particular wave requires you to reference something non-physical? Would you say waves are partly non-physical?