I"ve had something this experience before when converting from fumarate to freebase and doing various extractions. It seems like you can never get all of the DMT out of the aqueous layer, e.g. you only get ~80% on each pull, so it's 0.2*0.2*0.2*0.2…= 0.2^N remaining in aqueous layer for the Nth pull. I was never able to recover 100% of the DMT using this method, there would always be a layer of fluffy white DMT floating the top of the aqueous layer. Xylene seems to get more, but then you can't freeze precipitate… For this reason I now use sodium carbonate+acetone instead to do fumarate->freebase conversion, which seems to have less losses.