reasearch performance

hello
i am trying xindice as a possible solution for the storage of a 
dictionary, in xml format. my problem is that queries (the most 
important issue) are very slow, on my system (cpu G4 1,2 GHz) any 
queries takes about 30 second, that is a lot. i have tried with the use 
of indexes, and that helpes some queries, but there are some others (in 
the form of /somepath[contains(.,'term')] that don't benefit from 
indexes. there is any way to improve speed of these ones? o have tried 
also to split my documents (that were around 200 Kb each) in many 
smaller ones of 4 Kb each, but performance are even worst.
there is any evaluation of speed of queries of xindice avalaible?
thank you
filippo galgani