Well argued, and fair enough. I guess the question comes down to this:

and Lewis's conception of metaphysical possibility seems pretty close to mathematical possibility

I suppose I'm under the impression that the space of metaphysical possibility is much broader than the space of mathematical possibility. Would the two spaces be identical of a reduction if mathematics to complete and consistent logic had worked out? Does Tegmark take himself to be making an empirical claim, in asserting a multiverse?