When you're talking about executive performance at extremely large companies within specific industries in specific time frames, sample sizes are always too small to draw conclusions.

Then we are in agreement :-)

The thing is, small sample sizes, be it as they may, are still data, and drawing inferences off of them is the only thing we can do given the nature of our inability to experiment.