just a tip…..
Run it for 8 hours on the Pi, then plug it into a real computer for 8 hours and report back your findings,
I've see instances of 'A' shared being significantly down, even though the speeds of the miners match.
This might line up with what I was seeing. I mean, the cube on the Raspberry had full hash and got up to speed fast, but the number of rejects reported in the cube's web UI were way higher than when running through a PC. Eh, if I get another Raspberry I might just play games on it instead for a while, since it's in the mail already I guess that's what happens next...