sbt.. yeah.. it's in hybrid mode.  If you'll look back over that spreadsheet.. the changes to the materials I did was what made the large changes in the samples per seconds.  They did little or nothing with the kc/s or efficiency.. when i stuck the cube over everything that's when the efficiency took the huge jump.  Use the glass material being the major culprit of reducing render speed.

One thing, I am currently pricing out getting a new computer to do this and my cad stuff for school.  I knew lux's site had a benchmark chart for cpu's.  Lux's own measurement for how fast a computer is with luxrender is the contributions per second and not the samples per second or the efficiency.  What we might be doing looking at lux's efficiency and s/s speed is like measuring a car's fuel efficiency by how cold the air conditioning is.

I actually downloaded the luxtime benchmarking scene and ran it on this computer.  I was dismayed to find out that my computer scored 3 from the bottom on the test.  Looking back over that spreadsheet with an eye to kc/s.  Putting a box around the scene increased c/s almost 3 1/2 times in cpu mode and over 4 times more in gpu mode.