I'm putting this here because I was asked to do so (probably for completeness in discourse)
I have -never- used {le} to indicate a non-veridical description. I am completely on-board with selpahi's observation that nearly all human language is metaphorical and explicitly marking a description as non-veridical seems to be a waste of breath. In most situations, your speaker knows when you're making a non-veridical description.
The example I habitually provide is, imagine some men are hanging out by a water-cooler and some women walking past hear them making sexist remarks. One women isn't going to confuse the other by saying something like "Those dogs are disgusting". Does the listener really require such a front-and-center indication of non-veridicality? As far as I'm concerned the answer is blaring "no". We can have some articles for explicitly marking non-veridical descriptions but they should hidden away in an obscure cmavo that is there only to fulfill the promise that we can be as precise as we want to in lojban. Use {pe'a} is my suggestion.
{le} has always existed as a Definite Article for me. When I first started studying Lojban I remember someone remarking quite aggressively that {le} is DEFINITELY NOT the English "the". Since then, I've heard similar disgust at even the notion that they share similar semantics. The impression I originally was given was that "oh god, -nothing- from _ENGLISH_ could be useful for communicating jbopre". Since then I have understood that the feeling is more that the English "the" is simply muddied up the same way {le} is muddied up with multiple semantics.
Well then, that's fine because I've only ever used it for the major definition of "the" as a standard Definite Article and in the most general sense at that. From wikipedia:
A definite article indicates that its noun is a particular one (or ones) identifiable to the listener. It may be something that the speaker has already mentioned, or it may be something uniquely specified. The definite article in English, for both singular and plural nouns, is the.
This is -exactly- how I've used {le} in the past and how I describe it to newcomers and even those jbopre who claim to have no clue what it means. Contrast it against the definition of the Indefinite Article:
An indefinite article indicates that its noun is not a particular one (or ones) identifiable to the listener. It may be something that the speaker is mentioning for the first time, or its precise identity may be irrelevant or hypothetical, or the speaker may be making a general statement about any such thing.
To me, this is natural, intuitive and not complicated at all. It has been pointed out to me that certain cultures don't have this distinction and so it is not natural for everyone to think this way. But for me it is indeed incredibly intuitive. The most important semantic that {le} introduces is a signal to the listener that the referenced object can be identified either through context, preknowledge or otherwise. This is especially clear when contrasted against object descriptions where no referent can be identified.
What does this mean? I like to use examples:
Let's say a teacher asks a student "What is your favorite fruit?" the student my answer with an indefinite description, "Apples!". In this case, there is no relevant apple. It isn't a quantification issue. The student isn't merely saying that they {nelci ro lo plise}. There's no quantification taking place at all, since there is no referent to the description. There is no context or Universe of Discourse which allows the listener to identify one ore more apples that the student was referring to. But not because there arent any apples around - but because the description is literally indefinite.
Now let's change the thought experiment. We simply introduce a table where a selection of various fruits are arranged. Now when the teacher asks "What is your favorite fruit?" the student will invariably give a slightly different answer that aims at the same objective of communicating their favorite. They will say "The apple!". The description is now definite. It isn't just definite in the mind of the speaker. There is a real, objective and practical difference here. The listener now has a previously inaccessible capability to identify -a- referent to the description. We're not talking about accuracy on the part of the listener, or the vagueness or specificity of the speaker. We don't care if the listener gets it wrong because they are dumb, or the speaker isn't precise enough. We're talking about a substantial and mechanical distinction in that the formulation of the speech creates the potential for identification because the speaker has provided a definite description. To contrast how definite and indefinite descriptions create completely different linguistic circumstances regarding referentiality and identification that is completely unrelated to vague or precise speech (ambiguity) notice that regardless of the precision of speech, no indefinite description can ever be identified as having a referent. There is no context in which the answer "Apples!" is relevant to any specific apple. It doesn't even attribute the students preference to all apples; its orthogonal to quantification.
Further, the mechanical difference -has nothing to do- with 'coining' a creative label for which to refer to something. A nickname, impressionistic, utilitarian or otherwise non-veridical description that we create in order to -help- identification. That's a second order concern once referentiality and therefore identification is *possible* in the first place. Non-veridicality implies a definite description has already been made.
Some people will argue that this distinction isn't useful or claim to have a complete inability to fathom such contexts where the distinction between definite and indefinite descriptions is actually useful for communication. I find it -childishly trivial- to come up with examples where the distinction between the indefinite and definite descriptions of some 'kind' or platonic class is important for encouraging comprehension between two interlocutors.
:: A just returns home to B::
A: {coi}
B: {coi .i xu do se pluka le draci}
A: {uinai na go'i}
B: {ue .i xu do se pluka lo draci}
A: {je'a}
A: Hello
B: Hey. Did you enjoy the play?
A: No, :(
B: What? Do you like plays?
A: Indeed.
Now this example assumes that {lo} doesn't contain the semantics of {le}, or rather, we can specifically utilize the semantic of {lo} that is inaccessible to {le} by contrasting them. If we removed {le} from the lexicon it would take a more verbose description in one of the cases. Either "the play that you just saw" for the first appearance of draci or "plays in general" or the second appearance. By letting the gadri indicate or signal to the listener that the {draci} that we're referring to is definite, probably contextually relevant or in some cases perviously mentioned but more importantly identifiable. We then constrast our second description against the {le} version by using {lo}. Two speakers who speak my preference in lojban semantics immediate recognizes this second question as asking about plays in a way where no specific play is identifable. Therefore we implicitly and automatically know that we're talking about plays "in general". The discourse is efficient and clear.
I have argued that {lo} and {le} have the same exact lojbanic expansion, {le} doesn't affect quantification after all, it just makes this metadiscourse signal to the listener. But if I had to come up with something, if {lo broda} is expanded to {zo'e noi ke'a broda} then, in my world, {le broda} is defined as {zo'e noi ke'a broda jecu se dubyfa'i do'o} or something like that. Its probably wrong, but it just shows that the identity of the thing for which the description is given must be concretely identifiable.
I appologize if there is a lot of tautology here, but I'm just trying to drive home the impression I'm attempting to put forth.