I think such people believe they're saying something close enough to what they actually mean

I think they believe they're saying what they mean, but what you say is what's actually happening. More generally, I think you're talking about how people should act, and I'm talking about how they do act, which is making it sound like we disagree more than I believe we actually do.

This makes it sound like "clear to me" is a highly idiosyncratic criterion.

Yes, and I stand by that--even though, as you say, there is such a thing as objectively less ambiguous language. Which parts need to be less ambiguous in order to guarantee understanding varies from person to person. Even if the truth is so simple that people who aren't neurotypical require, on average, less ambiguity than neurotypical people do, that doesn't mean you can communicate with all of them exactly the same. Which parts you can fudge depend on the previous experience of the specific individual you're addressing. If I say, "Hey, did you see the less wrong thread about grading?" you would know what I meant, but someone unfamiliar with LW would be entirely justified in calling that sentence confusing. Someone who's, say, a weaver, might go off on entirely the wrong mental path.

By that logic, I think it's absolutely reasonable for a teacher above the first grade level to assume that all of their students are generally familiar with the mechanics of grading and potential penalties for late assignments. Of the two intepretations of the disputed statement, "you will receive a score of 0" is much more plausible than "you will be freed from having to do the assignment" to anyone who's been in academia for any length of time. I agree with you that the chosen wording could be, objectively, clearer; I do not believe there was sufficient reason to expect it would be misinterpreted that it was negligent not to be clearer. If I ask you to "make me a sandwich," I don't expect you to be tripped up by the real grammatical ambiguity and wonder if I would like you to put me between slices of bread. I expect you to go with the plausible choice.

I do agree with you that most people don't think very much about how they communicate. This frustrates me a lot, because it's something I think about a lot and which is important to me. But I don't find this fact as inexcusable as you appear to. Shortcuts and built-in amibiguity are a part of, as far as I know, every human language. In almost all situations, they resolve without conflict. Very few humans have ever had reason to consider that their use of language is insufficiently ambiguous. (When ambiguity does result in conflict, we assume by default that the other person is wrong, misunderstanding or stupid--which is not good or useful, but happens, and reinforces our belief that the way we communicate is fine.) Could most people benefit from learning to communicate better? Absolutely! But they don't know that.

Being upset that humans speak ambiguously is a little like being upset that humans shake hands. You're free to dislike it, and choose not to engage in it, but actually being unhappy every time it happens is going to get exhausting. A more productive thing to do in both cases would be to educate people about what's wrong with it and how they can do better ... in a communication style, of course, which they will find compelling and convincing, even if you wouldn't.

Which is why, if you give me a few years, I'll have a degree in this.

ETA: Skipped something I think was important: While I do agree that we should strive to be clearer in our communication, I don't think it's feasible or even a good idea to try to be as unambiguous as possible all the time. Thinking carefully about everything you say is difficult and tiring, and it takes a long time. Optimizing for unambiguity would be a sacrifice of mental energy and communication throughput which I don't think it's worth, given that the system really does work most of the time. In my experience, it's easier and more useful to try to optimize for the specific understanding of the person you're addressing, which may include being less ambiguous but almost certainly includes choosing particular ambiguities that the listener is likely to understand as intended. This still uses some mental energy, but not as much (at a guess, that's because it's a more natural thought pattern), and it doesn't sacrifice throughput because you're just choosing your shortcuts more carefully, not abandoning them altogether.