In the current implementation of the gnutella protocall, I believe anything that's not a-z, A-Z or 0-9 gets stripped from searches. Anything you could possibly use, besides boolean queries -which would just be turned into search terms-, is parsed out, probably because a lot of people do what you did and put quotes or whatnot in their query. This includes periods, plusses+, minuses-, exclemation points!, commas,, apersands&, astrixes*, brackets []{}, parentheses(), etc...

Adding searching options like boolean expressions or logical (+ -) operators would be cool, but until everyone adopts protocall v0.6 (which is still in developmental stages) these could only effect the searches that the program displays, not the way the search is performed. This means the searches won't be any faster and just as much traffic coming in, but the files diplayed will be filtered more at the cost of CPU overhead.