Given the status of these texts as not being published by JRRT himself, he was free to change his mind about stuff, even drastically . . . or to look at it another way, folks will look at the extant texts and give various opinions.
My approach is to give chronology (when
Tolkien wrote a given text) it's due to try to find (if possible) JRRT's latest know idea . . . that is to say, barring actual publication by the author, I tend to go with later idea over earlier (unless the later conflicts with something already published).
This reminds me of another thread where these very points were discussed. In general, I agree with you.
However, my position may differ in this one detail, that I do not consider it entirely out of the question that, had
Tolkien lived as long as a Numenorian, he would (in extreme cases) have subsequently brought out a few changes.
These changes, although not in direct contradiction to previous releases, would put the story in a very different light, like that weird thing about
the One Ring that originally just made
Bilbo invisible...
The fact that the original description of the effect (or non-description of the negative effects) is somehow strange and suspicious in retrospect, and led to a revision in the new editions post LotR, is finally explained very well by a trick also Middle Earth-immanent.
Nevertheless, it is evidence that one should not set published works as final fact, but rather view them as the version that has been handed down through the
ages from a copy or retelling of the Red Book, etc., which can make any serious, subsequent change theoretically possible.
In addition, one could also learn from Christopher that JRR now and then jumped back to one of the earlier versions after several changes of mind (no matter which topic it was about, but prominent there are e.g. the
orcs) also after advanced time and considerations.
Consequently, I would say the same: Yes, published works have the highest rank, then come in chronological order his unpublished versions.
However, one should always keep open the possibility that he would have ultimately still somehow decided differently (or yet again as before).
In any case, I wouldn't ultimately commit to a specific version on issues where JRR apparently wasn't definitively sure or ended up contradicting previously published works. There should be no problem in admitting different possibilities that JRR considered, regardless of the publication status and regardless of the time, even if these two "coordinates" of course influence the probability extremely.
Everyone may also like to determine these probabilities himself according to his own system, but an ultimate canon does not really exist there in my opinion, because the premise or definition of "canon" already excludes to consider (in this case quite) legitimate alternatives. At most, one could say that he thought this way or that way in the year XY...
But let's not get it wrong: I don't mean to say that everything is completely arbitrary and open to free interpretation, as is often unfortunately done anyway in film adaptations for various reasons.
In 99% of the cases we would probably agree, and in the remaining one percent I would say, yes, it is probably as your system provides, but I also keep open other possibilities - based on his texts and his way of thinking - of course as I interpret them, and depending on my knowledge of JRRT's works and statements, and aware that he has changed certain views over time, again based on his character and experiences. 😉
But anyway, that's certainly way too philosophical and nerdy for the topic of this thread and sub-forum.