Members of a species can breed with immediate ancestors and descendants, no matter what classification we put them under later
That's the idea. The very limited interbreeding occurred around 60,000 years ago (b/w H. sapiens and H. neanderthalensis). The species had already diverged, but were very closely related. Researchers are able to identify specific places in our genome that appeared after we diverged from Neanderthals, including gene variants for cognitive development. I'm quoting from the recent Discover article here (May 2011, "Meet the New Human Family").
That two species can interbreed doesn't mean they're actually a single species. For instance, golden retreivers can be mated with grey wolves. Same species? No, there are a host of behavioral and physiological differences. I don't mind sharing my quarters with a golden retriever but I know a wolf will want to spray urine around the place at a distance of about three feet from the floor. That trait can't be trained out of it. It will challenge me at a certain point in its development--the golden retriever not so much. Also, the wolf has a bite strength of about 1500 lbs per squ. inch, the golden about half that. These traits all point to separate species, albeit closely related ones--hence the ability to mate and produce a hybrid.
THe point is is that the term "species" is a cultural construct that resists an absolutely precise definition in the real world. We like the idea as humans of a nature that is finely ordered and easily categorized. But that's not how life works. Evolution is sloppy, imprecise and fuzzy around the edges.