Not sure about the point of this. One particularly bad suggestion about the way to do relative clauses is put forward as a bad way to do it and then the way we have always done it is recommended. This leaves aside the uncertainties of the usual way of doing things and any consideration of other suggested ways of doing them. A review seems to be in order.
The standard way of doing restrictive relative clauses is (as often in tp) with two sentences. While the order doesn't really matter, it is generally clearer to give the relative clause first and then the main clause. So, "The man who can to dinner stayed a month" is 'jan li kama tawa moku. jan ni li awen lon tenpo mun'. Strictly the 'jan ni' could be just 'ona', but the 'ni' gives the definiteness of the English (not necessarily a good reason). This works fairly well, but does not separate the cases restrictive relative claues from non-restrictive ones, which are merely a different sort of fused sentences, giving more information about an identified object, rather than being part of the identification. "Charlie Brown, who is a bad football kicker, has a round head": 'jan Sali Bawun li pana noka ike e sike musi. ona li jo e lawa sike'. Perhaps conventions help here, as there is no temptation to use 'jan ni' in place of 'ona' in this case.
In some deep version of a language, the information in a restrictive relative clause is built into the specification of the referent. Thus, it would be clearer if that specification were part of the NP referring to the object. This is generally possible (though sometimes messy) if the referent is also subject of the relative clause: 'jan pi kama pi tawa moku li awen, lon tenpo mun'. But, since tp lacks passives and other rearrangement techniques, it does not work when the reference in the relative clauses is oblique. "The man whom I saw at the store is still behind me." 'mi lukin e jan, lon esun. jan ni li lon kin monsi mi' does not convert directly.
Tepan suggests (and rejects) simply putting the relative clause in after 'pi' as a modifier: 'jan pi ona li kama tawa moku li awen lon tenpo mu'. He says this is bad style, but it seems, in fact, to be a violation of any of the grammars that have been proposed for tp. It also has obvious problems with telling the clausal 'li' from the sentential and so on. It does, however, allow oblique connections. The somewhat milder suggestion -- from Sonja, even -- would be to put just the predicate, with 'li' after 'pi': 'jan pi li kama tawa moku li awen lon tenpo mun.' It has the 'li' problem again and is again restricted to subject references. However, to the first point. if we drop the 'pi', we get just 'jan li kama tawa moku li awen lon tenpo mun' and ordinary and no problematic compound sentence. Of course, it will still not work for oblique references and it does not distinguish out the factor of identification from the merely informative content. An old suggestion from Mato (as far as I recall it) used 'pu', then a tp word in search of a definition but at that time clearly in the punctuation area, with either the predicate without 'li' or the whole sentence (for oblique cases) 'jan pu kama tawa moku li awen lon tenpo mun' ''jan pu mi lukin e ona lon esun li lon kin monsi mi'. The use of the new particle 'pu' allows for some new rules which might remove some objections to this format (e.g. the scope 'pu' runs to the second 'li', say).
Of course, simply adding the predicate without 'li' after 'pi' is very close to the standard way of building descriptions. The only difference is that, unlike standard constructions, DO continue to be marked by 'e' and PPs are not marked at all. This is usually a minor shortening and keeps some information that is lost in incorporation, the prerequisite for normal description building, where 'jan pi alasa soweli' can come from either 'jan li alasa e soweli' or'jan li alasa, kepeken soweli' (among other things, probably).
In short, we seem to come back to where Tepan does, with the two sentence solution and ordinary description building.