Back in the day (and maybe still), English teachers made a big fuss about the difference between restrictive relative clauses and nonrestrictive ones. Both were virtually sentences but for having a relative pronoun in place of one noun. A restrictive relative clause was a part of the specification of some term in a sentence, along with, perhaps, adjectives and the like, so it was adjectival, part of telling what the term referred to and so subordinate to the main sentence. A nonrestrictive clause, on the other hand, picked up on a term whose referent was already adequately determined and merely added some more information about it, less important than that in the main sentence, but otherwise parallel with it. In "The man who came to dinner stayed a month", the clause "who came to dinner" is restrictive, since it helps Identify the salient man. On the other hand, in "The man, who came to dinner, goes home each night at ten", it is assumed that the man is already adequately identified and the sentence then tells us two thing about him: that he goes home every night at ten and, more or less incidentally, that he came to dinner, the latter in the nonrestrictive clause "who came to dinner". In writing the old rules in America were that nonrestrictive clauses were set off by commas but restrictive ones were not. During the Eisenhower administration, at least, the restrictive clauses used "who(m)" for the relative pronoun and the nonrestrictive ones used "that." I gather that many of these things were different elsewhere and that at least the pronouns have reversed roles even in America. Still, the distinction is fairly clear, even if expressing it is not.
In the depths of a Montagovian grammar, the distinction is even clearer: the restrictive clause is part of the sentential matrix that underlies the term, whereas the nonrestrictive is merely another sentence conjoined to that in which the term functions: roughly "(the x: x is a man & x came to dinner) stayed a month" for the restrictive versus "(the x: x is a man) goes home every night at 10 & (the x: x is a man) came to dinner". (As a result, logically, if not rhetorically, the nonrestrictive clause and the main clause are interchangeable: "The man, who goes home every night at ten, came to dinner", while the nonrestrictive is not: "the man who stayed a month" may pick out a different man altogether.)
In tp, the situation is, at the most basic level, that there are two sentences joined somehow and it is a matter of interpretation (maybe eventually convention) whether they constitute a restrictive or nonrestrictive clause equivalent. So we can have 'x li kama tawa moku' and 'x li awen lon tenpo mun' in either order. The first x pretty much has to be 'jan', maybe with a 'ni', and the second x can be 'jan ni' again or 'ona' . At various times, various arguments have been made for one pattern rather than another as being more clearly restrictive Over all, the weight seems to have come down for 'jan li kama tawa moku; jan ni li awen lon tenpo mun' on the ground that you have to have the subject identified before you can say anything about home and the semicolon and the 'ni' show the tight connection (the second choice has 'jan ni' at the beginning, a colon and 'ona' in the second place, even closer to the logical form). But a clear case can be made for 'jan ni li awen lon tenpo mun: jan (or ona) li kama tawa moku' since 'jan ni' and the colon clearly say "a man to be specified later" The nonrestrictive case is generally assumed to be 'tenpo pimeja ali la jan li tawa tomo lon tenpo nanpa luka luka. ona (or even 'jan ni') li kama tawa moku', with the main stuff up front and the rest tagging behind. And, of course, when the term is subject both times, the collapsed version is often available. If that is fixed, the rest are pretty free for the restrictive case, whichever way it is done.
There is, of course, another way to do restrictive case: 'jan pi kama pi tawa moku li awen lon tenpo mun.' More compact, somewhat more familiar, but also a little less clear in more complex cases: "The man who hunted ducks with me last Saturday is in the shop" 'jan pi alasa waso pi poka mi pi lon tenpo suno nanpa luka tu pini li lon esun. In addition, all connections get reduced to modification, so that the 'e' of direct object on 'waso' is lost and the ordinary PP positioning becomes 'pi'.
At this point, the thought has occurred -- whether consciously or not -- to various people at various times, "Why not just put the whole predicate in as a modifier in NPs?", after all, we can stick in PPs without any problem and they are predications sometimes. This has resulted in cases of 'e' turning up unprepared for in NPs, with the usual lecture about how that can't happen. As Sonja has indicated, the problem is putting in just the predication, which looks for a while just like another modifier string -- until you get to the 'e' or PP -- rather than the whole predicate, with the leading 'li'. There is no problem with an 'e' in a structure with a leading 'li' or for floating PPs either. And there is not immediate technical problem with allowing this move. So, 'jan pi li alasa e waso poka mi lon tenpo suo nanpa luka tu pini li lon esun' could be quite grammatical. There are practical problems with it, of course: it fouls the lives of people who parse by finding the 'li' and dividing the sentence thus, though it doesn't seem much worse than compound sentences for that (although it can be confused with them, too). It is hard to further modify an NP once a predicate has been incorporated, since the later modifiers may get sucked into the predicate (another job for commas?): jan pi li alasa e waso poka mi lon tenpo uno nana luka tu pini (,)suli' does not look quite right. And, of course, if the predicate ends in PPs or is in an object position before PPs, a range of ambiguities arise that even commas may not save. Still, there are no serious objections (like that it introduces another level of recursion into the grammar or whatever), so it seems an option open to adoption.
However, it. like ordinary modification, only works if the the term in question is always the subject: I cannot use predication or modification to deal with "the man I went hunting with" or "the man that I hunted". To do so would involve inserting a whole sentence into an NP, something like 'jan pi mi li alasa e ona li lon esun', which ruins every grammar and does add a new level of recursion (A short list: 'pi' followed by only one word -- you can't make make this an escape, since the subject might not be one word, but then the 'pi li' is lost -- 'li' after 'mi' -- which you certainly can't drop without the whole predicate idea collapsing --, no prior referent for 'ona', since the NP it is to refer to is not yet complete.)
So, a general way of dealing with restrictive relative clauses always comes back to two separate sentences and some convention about how they are linked. Or to significantly changing tp. One possibility of the latter sort is to add true passives. This has been suggested for other reasons as well, but goes against an early philosophical commitment of tp. To be sure, that commitment -- to not allowing verbal escape from responsibility by burying the agent in an easily dropped PP: "Mistakes were made" by someone unnamed -- can be easily circumvented with indefiniteness "Some body mad a mistake" or abstractions "Mistakes happen" (no agent even implied). But still, adding at least two new function words (to passivize the verb and introduce the old subject -- though that might be 'tan' or the like) seems a high pice for what is a small gain. Notice that this move would not help for cases like "The man I hunted" or "The man whose brother is the mayor".
The other way to change tp would be to add a device for embedding restrictive relative clauses as a separate sort of unit. Back when 'pu' was a word floating around looking for a definition and was said to be some sort of punctuation, several people suggested ways to use it for this purpose, basically a 'pi' that attached a whole sentence into a modifier string. But another word would work as well. Another word would also be needed, to indicate what slot in the new sentence was occupied by the expanding description ('ona' will not work logically, for reasons already noted, but might be pressed into service). Again, the added complexity (and the new level of recursion) seems more trouble than the gains are worth.
But, even though we do not get a complete solution to one-sentence, modifier-included restrictive clauses, the simple devise of inserting predicates might constitute a gain and one that people have intuitively (or unconsciously under the throes of L1) used.