Chris, that is not actually what we did. Personally, I find indexing simplicial sets by inner product spaces to be unnecessary and unhelpful, and I've not coauthored any paper with such a construction. One can easily compare symmetric spectra in simplicial sets with symmetric spectra in topological spaces, and one can easily compare symmetric spectra in topological spaces with orthogonal spectra. I see no point in a hybrid. As a matter of detail, in defining orthogonal spectra one can perfectly well work with all finite dimensional inner product spaces, without choosing a universe, whereas the universe is needed to define the linear isometries operad used in the EKMM construction. It is nice to keep the S^V as they are: that makes generalization to G-spectra effortless, where G is a compact Lie group, and that works for both orthogonal spectra of spaces and EKMM S-modules.
I prefer eclecticism: the different models have different advantages. Here is an eclectic correct definition: a map of symmetric spectra (of spaces) is a weak equivalence iff its pushforward map of orthogonal spectra induces an isomorphism of homotopy groups. (Proven in the paper MMSS Chris cites.)
ps: I really don't like ``if you really must ...''. There are serious advantages to working in a model category in which every object is fibrant, and, related to that, both for theory and computations it is very helpful to have a clean zeroth space functor from spectra to highly structured spaces.