Have you tried it with 1mhz yet? You can get a 50%duty cycle 1mhz square wave from the 328 pretty easy, tho idont remember offhand its only a few lines of code
I would suggest making a (few) prototype pcb from batchpcb or similiar , making it as tight together as possible, will ground plane, and then check again for your overshoot, you may find its entirely related to the prototyping method you are using