My language's documentation is not finished yet. Maybe I will let you know when it is, but it will need some time.
The biggest issue I see in Lojban is that positions in relations are too arbitrary and too few. On the one hand it is impractical to learn the meaning of all words' components. On the other hand they are still to few and need to be extended by further "prepositions". Having more components, the same system of components for every word and marking components by further position markers like fa, fe, fi, fo, fu would solve that.
Some operations for example to form definite interlaced relative clauses are missing. Maybe I just didn't find them yet. On the whole every operation seems to have its own characteristics instead of everything following a more unified system.
There are some minor issues with spelling like that ' and . are used, but h, w, q are not. But on the whole there are much much more problems in natural languages than in Lojban.