The plastic isn't perfectly flat, but I don't think it would sway the results that much. What seems to be happening is the diffusion directing light up and other directions. To some degree the photons are still there, but travelling up rather than down and not touching the sensor. Some of those photons would still hit the plant assuming the 29" mark is well below the canopy level.
I should note for this test that the lamp takes up much of the area and is designed to be hung at around 8 inches. If I had measured from that distance I expect the difference would be 2-3%. A more compact lamp would have somewhat better results on plastic at the same distances. That's why I mentioned variance approaching zero as the canopy distance is lowered. But for those who do like to hang their lamp high the mirror would certainly provide a large benefit over white.
@Randomblame, about 2/3 of the space (all but the doors) are permanently covered in mirrors but I taped a few mirrors in place for the test so results were closer to what 100% coverage would be. Originaly I was going to do the doors as well but it didn't seem like a great idea as they are removable. If they were on hinges I would have mirrored them too. I suspect mylar being fairly directional would perform similarly so I'll switch the white on the doors for mylar when I get a chance.