They believe no such thing. Selection isn't required for allopatric speciation. Two isolated populations will naturally start to diverge just due to drift (random genetic changes over time). Given enough time they will inevitably become sexually incompatible.
"As far as I know, no new species have ever been produced, despite natural or unnatural selection."
Then let me direct you to: http://tinyurl.com/ridley2004speciation
"We begin with two distinct, but related species. The pollen of one is painted on the stigma of the other. If a hybrid offspring is generated, it is usually sterile ... However ... the chemical colchicine can often restore hybrid fertility. It does so by causing the hybrid to double its number of chromosomes ... Hybrids so produced may be interfertile with other hybrids like themselves, but not with the parental species. They are then a new reproductive species."
"Most garden varieties of irises, tulips, and dahlias, for example, are artificially created species. But their numbers are dwarfed by the huge numbers of artificial hybrid species of orchids, which it has been estimated are being formed at the rate of about 300 per month."
"Polyploid hybridization is also important in natural plant evolution."
Now, that type of speciation is a special case. It is not how animals speciate. But you are quite wrong in your quoted statement above.
"how come we take it for granted that humans, homo erectus, neanderthals etc. were all divergent species that had a common ancestor something like 50,000 years ago? Am I way off?"
Yes, you are, by an order of magnitude. Homo neanderthalensis is supposed to have diverged 500,000 - 800,000 years ago. http://dx.doi.org/10.1126/science.316.5827.967a ("The new findings also push back the date that Neandertals split from the human branch of the primate tree by 200,000 years—to 800,000 years ago")
"maybe the white->dark->white butterflies was a bad one"
Don't believe everything you read in AnswersInGenesis. This is still a good example. Or at least, so I believe on the basis of http://bostonreview.net/BR35.2/block_kitcher.php - "These are interesting alternatives to the familiar story, and the causal hypotheses they introduce can be tested in obvious ways: by examining the rates of larvae survival or by investigating nocturnal motions of moths. And this is what biologists have done."
But I don't have time to dig into the primary literature.
"scientists have been interbreeding literally thousands of generations of fruit flies and despite all selection pressures, artificial or not, were never able to produce a single new species"
Do you have examples where this has been the actual goal? I know Dodd (1989, http://eebweb.arizona.edu/courses/ecol525/readings/dodd_1989... ) tried this, but I don't think the experiment ran that long.
You will appreciate that it is much more effort for me to correct your poorly-informed naysaying than it is for you to naysay.