I can tell you that on my FW setup, the PAR decreased significantly when I increased the depth of water in that tank. I actually ended up needing to upgrade the light fixture since I was getting so much less light at the substrate. Went from 36w of t5no to 62w of t5ho. I did measure the PAR in that tank with the original t5no fixture empty vs "full" (14" of water in a 24" tall tank) and my PAR was cut almost in half. That's not solid evidence though...just my experience with a t5 fixture.
I feel like any increase in PAR with water in the tank vs empty would be negligible (if measurable at all) and likely due to reflections bouncing around through the water column off the glass and surface of the water. PAR values in an empty tank tend to be consistent...but put a sensor under water and the meter tends to jump around due to ripples on the surface of water and light reflecting through the tank.
I wouldn't trust empty tank PAR readings either. It gives you a decent idea of the output, but considering the clarity of water in a SW tank I doubt you'd get those same values in a running system.
Edit: Here's another link to the new orphek light...as well as a video showing the PAR measurement 36" below the light. The tank looks to be about 24" deep and they got a PAR measurement of almost 200. No way that light is putting out that much PAR (~150) in a 7ft deep tank...as they alluded to on that other page. They light does look pretty impressive, but yea I wouldn't trust those previous numbers.
[video=youtube;9D8bR66Je_0]http://www.youtube.com/watch?feature=player_embedded&v=9D8bR66Je_0[/video]
http://orphek.com/led/orphek-products/atlantik/