I'll give a few options that occur to me.
- Talk to the editor. Essentially ask him what to do. On the one hand he has a policy he wants to follow, and on the other hand he presumably wants corrections of erroneous papers to appear. After all, the alternative is that someone points the error out in a different venue, and the whole community laughs at his journal. Perhaps he will be willing to bend his rules a bit.
You will need to discuss any possible solution with the editor, anyway, so you may want to start out with the most obvious option.
What's to keep you from writing something titled "Erratum" with the whole list of authors X, Y and Z, but with two sections "1. The Erratum (X, Y)" and "2. A dissenting point of view (Z)"? Yes, of course we want a paper to have a unified message, but in this particular case, I'd say that getting the word out in any way is better than burying the discussion under mounds of policy.
Propose that you write the erratum, but allow the dissenting author to comment on the erratum and explain why he thinks the erratum is erroneous. Then send both the erratum and the comment out to review and see what the reviewers say.
Perhaps you can do the discussion under point 2. above on the ArXiv, but publish an "erratum" with the full author list that does not discuss the error but only points to this discussion on the ArXiv.
One bonus for options 2-4: actually writing up these papers will force everyone to really think about the derivation in question, and either you or the dissenting author may change his mind about whether there is or is not an error.
Option 1 puts the onus on the editor, who may not be very flexible, judging from your question. (Of course, we don't know what he would say.) Options 2-4 are pretty "meh", and involve getting both the editor and the dissenting author on board. So given that these three options involve convincing two people, anyway, it may be worthwhile to look at other solutions involving communications.
How deeply have you discussed the putative error with the dissenting author? It may be easier to convince him that there is an error than to convince both him and the editor to collaborate on a "solution" nobody is happy with.
How sure are you that in fact there is an error? I'd assume the dissenting author is not completely out of his mind, so perhaps he does have a point?
Options 4 and 5 really involve taking the time to discuss the issue in depth with the dissenting author. I'd recommend doing this in person, if at all possible, not via email or the phone. And do this one-on-one with the dissenter, not together with all the other authors - if he perceives everyone else "ganging up" on him, he may become defensive.
Approach this with an open mind. Maybe there is a simple misunderstanding, and the two of you are really thinking of different things when you think you are talking about the same thing. Often you will learn a lot from this kind of discussion. In the best possible case, you will find two different ways of looking at the issue, and maybe even the germ of a follow-up paper.
- If all else fails and the editor is in fact an associate editor, you can always escalate the issue to the chief editor.