Modal logic is actually quite useful. If modal realism turns you off you can just accept it as a language game (which any sort of formal logic is going to be.)

The non-sequitur in Plantinga's argument, as presented by cousin it, is P3. (Plantinga's own argument is a bit more subtle, and its ultimate error is in eliding between different meanings of the term "possible." He successfully shows that under formal logic if possibly necessarily x then necessarily x, and then ascribes possible necessity to God because God is one of the most few things that often is argued to be necessary, and that God seems like the sort of sufficiently abstract thing that it might be necessary. But this isn't the sort of possibility that's germane to formal logic.)

I understand that when folks say "modal logic" in this context, they're generally referring to model logics that implicitly quantify over poorly-defined spaces. However, that's not what all modal logics are like, and so I hate to see them maligned with a broad brush.

Consider, say, dynamic logic, which I actually use as a tool in my research on program analysis. When my set of "actions" are statements in a well-defined programming language, I can mechanically translate any dynamic logic statement into a non-modal, first-order statement. I almost never do this, because the modal viewpoint is usually clearer and closer to the way we actually think about programs.

Equivalently: you can use whatever logical operators you like, if you can define the operator's meaning without reference to the operator. It can help you say what you're trying to say, rather than spending all of your time with low-level details. It's like a higher-level programming language, but with math.