This is a cool problem, and it _is_, indeed, quite hard! I hadn't seen it before.
If we look at the primary solution offered in the Wikipedia article, it does not have the fatal flaw you think it does. The solution does assume, as you say, that "one of the gods must answer a factual question truthfully". That's fine, as that is part of the statement of the problem: "True always speaks truly", and we know that True is one of the gods. But nowhere in the rest of the solution is there the conclusion or assumption that "ja" corresponds to "yes" and "da" corresponds to "no." It's an intricate and complicated solution, and so it can be hard to follow, and I totally empathize with perhaps misreading it.
I don't think I can write up in this space another presentation of that solution that's better or clearer than the one in the Wikipedia article to help show that it's never assigning a specific meaning for "ja" or "da". But if you'd like to schedule a session with me, I'm happy to walk through the solution together and show you how it works and address questions.
I can't really help with questions about whether there are harder, unsolved puzzles. There are plenty of unsolved, interesting mathematical questions out there, but I'm not sure they're puzzles, per se. You might be interested in Hilbert's problems, if you're not already considering those: https://en.wikipedia.org/wiki/Hilbert%27s_problems
Your questions 2 and 3 are basically answered by the Wikipedia article, if you accept that it actually does solve the puzzle: it's a valid solution, and it even has a truth table!