By the way, I've always preferred to capitalize heaven and hell. Editors have usually lowercased them, citing a similar principle to the one we're discussing--in less technical (that is, I guess, non-theological) contexts, they're supposed to be lowercased. I deviate on that one.
I've never been happy with lower case for something like "the president," when that's a substitute for a particular name. That one I follow because it's so generally done that way. But I've never liked it.
I've got a tendency to split the baby with my editor on this kind of thing. I use Heaven if the context is religious in nature, heaven if talking about the sky. Generally, when using profanity, hell doesn't get capitalized, although I'm happy to make an exception if someone utters "I'm going to blow you to Hell and gone."
For ranks and titles, generally, if I put "the" in front of them, I don't capitalize. This is one that I adopted from my editor. In dialog, if I'm using someone's rank in lieu of their name, I tend to capitalize. The president gripped the podium until his hands turned white... versus, "Mr. President, you can't mean it..."
All of those usages are consistent with style guides in general, and that's pretty much the pattern I follow as well. I was just pointing out that some of them aren't
logically consistent. Grammatical rules aren't entirely logical, and they do evolve over time.
Most of my editors have wanted to follow CMOS to the letter. In fact, when I used to use CreateSpace editorial services (back when they existed), the editors noted the section in CMOS where the rule they were invoking was located. Later, I had an editor who advocating breaking with CMOS on some issues. For example, he wanted to hyphenate words for clarity which CMOS said should be closed (one word, no hyphen). That seemed like a sensible way to go, but I had to retrain myself to be consistent to the more subjective clarity rule, rather than a fixed "Prefix x is never hyphenated" kind of process.
That said, I suspect that if I were submitting to a trad publisher, I'd format using CMOS unless it was utterly ridiculous to do so, just to play it safe.
I've also had to retrain myself to accept certain CMOS usages that vary from the grammar I used to teach from. For instance, in that text, coordinating conjunctions only worked within sentences, not across sentence boundaries. That meant in practice that you couldn't start a sentence with
And or another coordinating conjunction. The current edition of CMOS actually ridicules that rule (though it does also note that it's still taught by 50% of English teachers). The grammar book usage seems more logical to me, but the CMOS alternative does have the advantage of making it easier to control sentence length without having to overuse simple sentence patterns.