Let's stipulate that the difference between "no god" and "perfectly uninvolved god" is bigger than any difference between different theistic scenarios. Would that really do much to resolve our disagreement about how to explain the survey differences? I don't think so.

The size of the difference has absolutely nothing to do with my point. My point is that the difference is of an entirely different kind.

To take an analogy; let us say you have a duck, and you are measuring the greyness of its feathers. This runs along a spectrum from snowy white to ebony black. There is no point on this axis where the duck is actually a swan.

the way you've phrased #3 reduces your argument to a tautology

No, I don't think it does.

Both of your examples appear to me to show someone, having changed their ideas about #1, in the process of altering #2 or #3 to match. On consideration of this, I will admit that I was thinking only of the steady-state case (when someone's beliefs are internally consistent) and not really thinking about the transitional period during which they are not (even though some people might spend a majority of their lives in such a transitional state).

I don't think the god-switch (#2) is so easily flicked.

It's not so much that it's easily flicked as that it has less moving parts; flicking it requires adjusting one thing as opposed to many things. (Of course, changing either can be difficult - the default reaction would still be to reject any unwanted evidence and/or associated arguments).

...though I could be wrong about that.