1st paragraph: Yes, my IC ballasts are the newer style. As for reflectors, I do use the dreaded PFO mini-pendants due to their size, but my results were not taken with a reflector. I made a test bench, er, black box much like sanjay uses. Its a 2' plywood cube painted black inside with a 250wattDE socket, and a mounting frame for the Tempered LoE glass which I borrow from one of the PFO's. The tests are best done w/o a reflector, as Im sure you are well aware... the positioning of the sensor under the bulbs can vary the output readings of the bulbs wildly depending on how the reflector's fascets converge the light. I didnt want to say it, as I believe your readings overall are pretty close to what I was getting, but to test the bulb output over time with a reflector makes it hard to be sure of the accuracy because a slight change in sensor positioning can change the results. The reflector changes the distribution of the light from a pure point source. Perhaps this is why your readings were slightly higher than mine... maybe the sensor happened to be under more of a 'hot spot' for the more recent readings?
2nd paragraph: no reflectors... just pure point sources in the 'black box'. Having the sensor 2' away prevents optical problems as well (even without reflectors, the bulb still has certain 'hot spots' just from its own glass... so to eliminate this, taking readings from 2' away helps.
3rd paragraph: you said, ...'and slight "under driving" will help it last longer'... This is not true. Underdriving is just as damaging if not worse than overdriving. The bulb's contents dont fully excite, so some of the contents just sits in the bulb, not giving off photons. While in this state, they burn into the wall of the tube... so you arent only losing the output from those atoms, but also blocking the output from the atoms that are excited because those elements are blocking the output. This is why I have taken PaulErik's advice to heart, and only run HQI rated ballasts with HQI rated bulbs. He is the one who first told me about the dangers of underdriving.
On a side note: Your reflectors could be swapped out for something more like a lumenarc DE, and you would see better penetration into the water than what you currently have. That target range of 400-500 is right on the money from what I have seen as well (havent seen anything that can even tolerate much more than 500), and many people are doing well with high-light corals under much less. I would almost extend that 'high-light' category to be as low as 300, as long as the coverage of the light is over more of the coral tissue. I have seen high-light milliporas color in better/grow faster under T5s in the 300 range that other wise needed light in the 500 range with halides, and I can only attribute this to the T5s being able to hit more of the coral's surface with light because they are more spread out as a source.
Here, I have been taking photos and then combining the light levels from the meter into composites at the WRS forums...
http://www.wisconsinreefsociety.org/...topic.php?t=87
http://www.wisconsinreefsociety.org/...topic.php?t=41
http://www.wisconsinreefsociety.org/...topic.php?t=44
You might recognize a couple there...lol. The results have been very informative, and provided people with exact info on how to better their systems. The meter for $318 was money well spent. Its given me more useful info than the $1200 spectrometer has at least... so far.