Marcus, the point with E8, and much of the amateur approximations we do here, is that from the point of view of high academia they fall in the category "tried without success, not PhD will be produced from here". Or, Also near of Feynman's analogy about boxes and keys, "did you try this number?".
The problem is that one must have a natural method to break E8 into the standard model. (And Garrett wants gravity too!). In the same sense that string theoretists have a lot of vacui to choose from, GUT theoretists have a lot of breaking patterns to choose. The only fundamental (*) advantage of string theoretists over GUT theoretists it that the later do not have a fundamental reason to break the group at all, while the former must do it to go down to four dimensions. BTW, Lumo uses this fact to claim that "Standard Model has been proven to be a consequence of compactified heterotic string theory back in 1985, I mean by Strominger Horowitz Witten Candelas". Fortunately the preprint of this paper is readable online http://ccdb4fs.kek.jp/cgi-bin/img_index?8504007 so you can read and check. As far as I remember, four 27-generations with N=1 supersymmetry and E6 gauge group is not the Standard Model. Yet.
(*) they also have to have one technical advantages, better control of V+A interactions, and one half-fundamental one: justification of the high energy cutoff.