Is the sentence you offer:
All circles are not unround.
nonsense? This is a good question.
The first thing you need to understand is that language is conventional. What that means is that what makes sense and what is useful and what language means is dependent on the context and the people using it. Consider how the sentence:
Uijt jt b djqifs.
seems to be utter nonsense at first glance, but might turn out to have meaning if it's a cipher of a single rotation. See how sense and nonsense are somewhat tricky? In fact, this notion that sentences might feel meaningful in the absence of a literal truth is famously enshrined in Chomsky's "Colorless green ideas sleep furiously.". So, I think it's fair to say that your sentence is mostly sensible and meaningful. The only part that is really challenging is the word 'unround'.
What does it mean to be 'unround'? Normally, school children run to a dictionary, which is full of what philosophers of language call lexical definitions. Does it have an entry as a reflection of use. In fact it does! (MW) But reading the definition quickly communicates that the sense offered here has to do with phonology and not geometry, so that gets us to the question, what do we do with a word that seems meaningful but doesn't have a dictionary entry? Well, if you're astute, you'll note that dictionaries are relatively modern inventions; people have been using words meaningfully long before a brave Scotsman started the OED.
One tool the philosopher has in his quiver is the stipulative definition. Here, for the point of the conversation we can simply stipulate that 'unround' is defined as 'not being round like a circle'. Here the problem is solved. Now, a square is unround by definition. And not unround would on the classical notions of logic (the law of excluded middle and principle of bivalence for instance) would be the opposite of unround which is round. So, we can see that 'not unround' can be transformed grammatically to 'round'. In the context of your claims, everything would seem to work out.
But what you are asking is, can I just stipulate definitions? If I do it, then anyone can do it right? And the answer is yes. In fact language use is highly democratic in this way. Countries often try to create language academies to regulate usage, but in the end, languages and dialects change. Modern English for instance differs from Middle and Old English which is related to Modern Frisian, but both part of the German families of language that are a branch on the PIE tree. It's hard to stop people from using language however they see fit, and that brings us to Wittgenstein and the notion of the language-game and the struggle of language prescriptivists who are constantly trying to get people to use a grammar according to a political agenda.
That brings us to the answer of your question. Like MarcoOcram noted, what you are talking about isn't a question of logic, so much a question of language. The use of 'unround' would be considered non-canonical or idiosyncratic or some would call it idiomatic. This is the idea that the language as presented is non-conventional. But on the whole, I'd say it's fairly meaningful, and I think 99 out of 1,000 people wouldn't have a problem responding to this as if it were conventional.