I don't really understand your first sentence. For instance E7 * G(standard model) is still unsuitable for phenomenology even if you are deadset on trying to complicate the high energy landscape for no suitable rewards!
Actually, the problem with SU(N) with N finite, is actually a little more complicated than I said above. In order to have a physically viable candidate, you need both a complex representation (and hence a chiral spectrum), which SU(N) does have, except that you also need to satisfy the additional requirement for anomaly cancellation. The only completely antisymmetric m fold representation for SU(N) then only contains the familiar SU(5) representation 1, 5, 5bar, 10bar. Which you can build up to get higher versions.
Unfortunately there again you run into certain choices for some fixed N, with a fixed representation r where you will have to explain why either fermions don't receive large masses or why the theory doesn't possesses possible gauge anomalies..
For N --> infinity, well that's a different story b/c of the aforementioned simplifications. But then that's not a candidate GUT either.