We actually adjusted our algorithm in April because it was OVERreporting scores in the verbal section (giving people higher scores than they should have gotten).

The test does NOT score you on # right / wrong. The algorithm relies mostly on the difficulty level of the questions you answer. You can get about half the questions wrong and still score a 700.

Also a 40 point change is within the standard deviation (about 50 points) - so that is not a statistically significant change (unfortunately, since it 40 points up!). Nor is it strange at all to see that kind of change (up or down) - that's exactly what to expect, actually. The official test has a standard deviation of about 30 points.