Hmmm. I may be using "metaphysical" inappropriately here. I confess that I am currently reading something that uses "metaphysical" as a general term of deprecation, so some of that may have worn off. :)

Let me try to answer your excellent question by analogy to geometry, without abandoning "metaphysical". As is well known, in geometry, many technical terms are given definitions, but it is impossible to define every technical term. Some terms (point, line, and on are examples) are left undefined, though their meanings is supplied implicitly by way of axioms. Undefined terms in mathematics correspond (in this analogy) to words with prior metaphysical meaning in philosophical discourse. You can't define them, because their meaning is somehow "built in".

To give a rather trivial example, when trying to generate a naturalistic definition of ought, we usually assume we have a prior metaphysical meaning for is.

Hope that helped.