Don't leave us hanging!
But on simple inspection, the idea that a word can fit into a single category is obviously wrong. So to generalize the idea, you have to assume that a word can fit into multiple categories.
Biology is undergoing a similar transformation from simple "this animal looks kinda like this one" to genomics classification, which is revolutionizing the tree of living things used by biologists to organize living things.
There's not really any reason why words, a biological artifact, have been correctly classified all along. Hell! A copy-editor neighbor and I have minor quarrels over the capitalization of college subject names.
Trees are a specialization of graphs, and they're usually wrong. Just assume graphs to start with and your data will organize much better, even if the properties of the organizational metaphor aren't quite as nice.
But I don't have any real reason to claim that they don't work. Syntactic theories aim to explain why particular sentences are grammatical or ungrammatical (in the judgments of native speakers). If you can accurately do that, then yes, you've put each word into one category (as far as making up grammatical sentences). So to say the kind of thing you're saying, you'd have to explain why the current research programs in syntax are misguided.
Ask a simple group of 5 linguists and they all will give you different answers. This literally happened in a workshop we gave at a college months ago. Quite funny.
While working on grmmr (green-bridge.org) we recognized that so many of us are densely confused when it comes to these metalinguistics. So we just stopped using knives that dissected (metalinguistics) words into words and went for friendlier marker (symbols and colors).
We need to be reminded that a lot of it is just pretend and educated guesses, no absolutes. Markers seem to work better here than knives do when teaching and categorizing language grammar.
The real problem is that taxonomic classification is rarely correct in nature.