This seems like a good time for the sorta annual reminder thaat Lojban as currently constituted is doomed to fail of its intended goals. The reason is simply that it is built upside down, so the proofs all have to run up hill. This pattern was set by JCB on day one, when he decided to work by adding logic to a speakable langauge (English), rather than extracting a speakable language from loigc (not FOPL, the then favorite, but, as should have been obvious already in the early days of Loglan, what is now called Higher Order Intensional Logic (HOIL)). We know this can be done because it is a given in (certain, e.g. Montague’s) linguistic theories that that is how languages actually come about.
I agree that Montague's work is worth a look, but it's hardly fair to fault JCB on that point. JCB developed Loglan between 1955 and 1960 and Montague's work was not published until about a decade later. In fact, due partially to Montague's premature death, and partially to the baroqueness of the formalisms Montague employed in his terse papers, Montague semantics (and what became formal semantics) did not become well known and understood for yet another decade after that. So to criticize JCB on that point is more than a little anachronistic.
(By the way, sir, forgive me for saying this: Given the timeline and given the fact that JCB was not a logician, it has always seemed to me that if *anyone* was in a good position to incorporate Montague's work into the development Loglan/Lojban, it was plainly you!)
Since we're on the topic, I will add that, in my opinion, Montague semantics in its original form would not really have been a silver bullet for the semantic issues and debates that have cropped up over the years on the Jboske list and elsewhere. Montague had no inkling of either mereology or plural logic, and seems to have bought into Russell's quantificational analysis of definite articles (which I think is much more aptly modeled as a Hilbert-type choice function, but I won't get into here). There is also inherent in Montague's work a deeply problematic conflation between genericity and intensions, and correspondingly between specificity and extensions (which I also won't get into here). The point is: the Great Gadri Debates probably would have happened anyway, simply because the issues themselves are tricky.
To be clear: This of course is all 100% Monday-morning quarterbacking on my part. I consider Montague as something of a genius, and I rank his contributions to the field of loglanging as being in the top five, if not top three of all contributions. But the point is that Montague's papers were not destined to save Lojban from semantic confusion even if they had been known about, which they weren't.
In summary, I'll say it seems pretty obvious to me that Lojban/Loglan does not do a very good job of being a loglang -- but it's only by *modern standards* I say that. Considering that JCB was (if I recall correctly) a psychologist working (as far as I know) all by himself from 1955 to 1960 with little training in either linguistics or logic, I actually consider Loglan pretty good from that perspective. It is most likely better than I could have dreamt up or built if I had been alive at the same time. So JCB takes his place as a bright star in the constellation of loglang history, regardless of whether his language fails or not to be a good loglang.