In fact Phoenix just makes it run too hot without noticeable speed benefits (using sdk 2.4).
Also core 0 is always about 5% slower than core 1 even though both have the same exact flags.
I think it stands to reason that if the choice of software is causing the card to run hotter then it's likely getting your card to do more work. The MH/s in POCLBM might just be measuring/displalying the hashrate while hashing so without all the network overhead.
I'd try running a benchmark by looking at the number of shares done by each core over say a couple of days.
I'm doing this exact thing between phoenix/poclbm and phoenix/phatk+mod