It is possible that new alelles can form without a mutation even occurring, so Faith's concept isn't entirely wrong. I don't think it has been shown to be a major source of new alelles, for reasons that might be apparant.
From an encyclopedia article on
recombination:
Recombination was once thought to occur only between genes, never within them. Indeed, the supposed indivisibility of the gene was regarded as one of its defining features, the other being that it was a single unit of function. However, examination of very large progenies shows that, in all organisms studied, nearly all functionally allelic mutations of independent origin can recombine with each other to give nonmutant products, generally at frequencies ranging from a few percent (the exceptionally high frequency found in Saccharomyces) down to 0.001% or less. Recombination within genes is most frequently nonreciprocal.
Some of that isn't germane but the first part is. To simplify it has been said that recombination is the reshuffling of genes. Given where Faith is coming from I think a little more is needed (sorry if it has been brought up already).
There is no process or marker which says 'this is a gene' in the DNA. So it is possible that during one part of meiosis (where the DNA is sliced up and recombined) that a gene will be sliced up. The shorter the gene the less likely that is. If a gene gets sliced up and paired with another sliced up gene with some other purpose, it might create an entirely novel gene.
I can't see any evidence of this actually happening, but I might be wrong. Dawkins goes to great lengths to talk about genes as a unit of selection, and one selection pressure is for genes to not be too long, or they will basically be destroyed with this method.
I'd be interested to hear from Faith and the biologists for comments on any studies or the like that have been done in this area. I believe that this kind of thing might technically known as a mutation, just as inversions and deletions are.