presumably running the usb code on the chip alongside the sketch complicates things to much.
Having a seperate chip means for people who dont want to deal with that extra level of complexity/customisation can just carry on as before (as can all the code already knocking around). For those who want to dig into those things solutions will emerge. Dean who writes the LUFA code running on the usb chip is already mentioning arduino specific example sin the next release of LUFA