Ok, so I calibrated the app and dunked the phone. Out of the water it seems stable and repeatable within 10 PAR or so. Every now and then I would get a wild reading, but would just remove the phone and put it back under the test condition and it would come back to the expected value.
Now to the tank... The ziplock bag that I used I think caused more problems then good, so much so I almost just dunked the phone in the tank. I've actually put my phone underwater before so I wasn't too worried, but I didn't go that far. The plastic bag kept triggering the HOLD feature on the app so it wouldn't update the reading. I would have to take the phone back out of the water and move the bag around for it to go live again. That was probably the biggest pain of the process.
As for readings.... I have a Radion XR30 Blue mounted 7 inches above a 30x30x24 cube in the center running at 50% intensity on the AB+ preset. At a depth of 8 inches directly under the light I had reading that ranged from 250-410. At the far corner bottom of the tank I had readings of 170-250. So I'd like to think that the low readings were the "correct" reading and that the plastic bag may have caused some abnormalities but who knows. If that's a ballpark reading, that's a pretty big ballpark. To my eye, I think the reading from the app is hot, but I can't be sure till I actually pick up a PAR meter.
This got me thinking... are there known models of light dispersion for these lighting fixtures? I would assume light transmission/falloff through air/water would be a constant... like an XR30 at 10" above the water at 100% intensity would yield X at a depth of Y. In computer graphics, and I'm sure other disciplines, we use IES files which model the physical light distribution and falloff of practical lights. I would love to see something similar for reef lights.