I have a vague memory of e-mailing Dawkins a decade or so ago about group selection and getting a response which more or less summed it up to my satisfaction: There's evolution of evolvability (or something like that, he had an interesting phrase for it), which is to say, group selection can take place based on individual-level selection pressures. The example, IIRC, was the tendency for certain kinds of species to grow larger with longer reproductive cycles, then go extinct as their reproductive cycles extended out to the point where they couldn't evolve fast enough to keep up with changing conditions. Other types were individual adaptations whose dispersement gave their groups massive advantages, outcompeting all other groups; the example there, IIRC, was sexual reproduction.

Which is to say, it's wrong to say that group selection doesn't exist, but it's also wrong to say it trumps individual (or genetic) selection. Rather, the entire concept of "group" selection is wrong in something the same way "individual" selection is wrong, because it is genes, in the context of other genes, which are selected.