You're not specific about which fruit fly experiments you're referring to
Probably because no such experiments exist. There are lots of mutational experiments on flies in the literature but I have never come across any where the purpose was to produce beneficial mutations or create a new species of fly.
Mutational screens are designed with the intention of detecting mutations with effects on the phenotype and almost all of these will be deleterious, because it is far easier to cause significant damage to a genetic system by random mutagenesis than to improve it and easier still to detect many such damaging mutations. It is doubtful that any beneficial mutation would present itself in a highly visible way from simple morphology, unlike a large number of deleterious mutations which can produce highly dramatic morphological changes.
So it is unsurprising that mutational screens fail to find what they aren't designed to look for.
I don't know what creationist site it is that sells people on this rubbish argument, but it is so prevalnet that it has to be coming from some common source, I can't imagine so many people thing up such a stupid argument reliant on complete ignorance of the subject on their own.
The classical mutational screen would be the Nusslein-Volhard and Wieschaus screen for embryonic lethal mutations in drosophila, which the astute will realise was designed to detect mutations which were lethal to the organism while it was still embryonic. I'm not sure how creationists expect such screens to detect beneficial mutations.
TTFN,
WK