I don't think this is in essence any different from various ontological arguments; I don't think it bears much relation to Plantinga's Evolutionary Argument Against Naturalism. (Plantinga does have a version of an ontological argument.)
Plantinga's ontological argument, and your argument, manifest a confusion about the difference between what one can imagine (or state) and what is real. That is, you start by granting that our minds can outline states of affairs that are not real, model worlds if you will; and then you insist that because some quality or property or proposition is true in the model world, it has some bearing on the real world. It need not! We can formally state all sorts of things that have no bearing on the real world at all, and in at least some sense we can imagine them.
There is no real avenue for improvement because it is exactly this (invalid) jump that forms the crux of the argument. In fact, pretty much every premise of yours is highly doubtful in the sense you mean it once this jump is ruled out. "All conceptions of reality are derived from the natural world" may be true causally, but only in the sense that all YouTube videos are "derived" from nonlinear properties of electrons in doped silicon. It's not the kind of derivation you need to conclude anything about the real world given our models of it. To show that strong kind of derivation you would need to undertake an extensive empirical study, and would probably come up empty.
Likewise, conceptions of reality need not, prima facie, be a coherence of meaningful experiences (though to call it a "conception of reality" we would expect to find some coherence and meaning even if fragmented). It seems doubtful that people believe that God shares no properties with the natural world if you allow that things like "love" and "justice" can be statements about very complex processes in the natural world. People extrapolate all the time, and P5 basically says, "No extrapolation, that's cheating!"; also, advanced pure mathematics is almost entirely about stuff that is meaningful in contexts absurdly disjoint from properties of the natural world.
(For what it's worth, Plantinga makes a very different kind of mistake with the EAAN, which is to confuse a possible world that he can imagine with the actual world and its correlation structure; in cases where correlations are unreliable or biased, we do in fact have unreliable perceptions, e.g. passage of time when one has a fever, all sorts of impressions about our internal state, etc.. The EAAN is a good argument that not any natural world could be comprehensible to evolved organisms, but it is a horrible argument that our natural world doesn't reward comprehension with fitness.)
So--sorry!--this sort of tactic doesn't generally lead anywhere (hasn't ever yet, at least), even if it can be transiently interesting along the way.