I am inclined to conclude that a certain uncertainty is built into Esperanto, and this isn't all bad.

The usage is a fact. The description is an attempt to account for the facts. A bad description is dangerous in so much as it may then distort the usage or the understanding, and if it presents Esperanto as being more complicated then it actually is, then it is disadvantageous from a propaganda point of view.

All languages, I would suppose, have the potential for ambiguity, though Esperanto does a good job in limiting that. A proper description may expose a potential ambiguity, but this is good in that it shows us how to be more circumspect in in our use of the language.

What isn't good is a false theory that incorrectly prescribes that a certain word must mean X and can't mean Y, or that a word, that has been used without problems throughout the life of the language, is 'evitinda'

Saluton! Just gonna inject myself in here and say that I really find myself agreeing with Nornen. I might have totally misunderstood everything that's been said, but it seems like the real problem is that there are actually two subjects being discussed at once. Sinjoriĉo Sudanglo seems to be speaking of how it should ideally function - how it should be structured to be the most logical, while Sinjoro* Morfran seems to be concerned more with how it does currently function according to the vast majority of speakers and the overall current scholarly position. This is why I agree with both of you as well. I personally prefer the idea of roots only having the class given to them by their suffixes (-o, -ig-, etc) but it seems... unrealistic... to expect to change what has become a core aspect of the language as it's actually used by millions of people - especially by fiat, and even moreseo by the fiat of a few people on the internet. It would have probably been better if the idea had have been incorporated into the language from the start. I can agree with that, whatever my opinion as a barely competent beginner is worth. But what's done is done. The time for doing this was about a century ago, and the people with the influence to make these kinds of decisions about basic grammar with any real hope of getting the parolantaro (is that a word?) to adopt it en masse so are long dead.

*Using it sex neutrally here, with a mr./ms. meaning as the case may be.

Sudanglo seems to be speaking of how it should ideally function - how it should be structured to be the most logical, while Morfran seems to be concerned more with how it does currently function.

I can't speak for Morfran, but I am saying the description should be congruent with the facts of usage, and that the idea that the derivation comes from the meanings of words is a better description than the idea that the roots themselves are of fixed grammatical class.