The only reason I can come up with for not using the "point it perpendicular to the direction of the target" approach is that it may give you a less precise reading. While intuitively, it makes sense that for any wind direction the velocity is composed vectors parallel and perpendicular to the target; in a perfect world, pointing the unit perpendicular to the wind direction should eliminate the parallel vector and give you a wind speed for the perpendicular vector you actually care about. But, my feeling is that the units are not sufficiently perfect that a wind 90deg to the unit does not spin the impeller. As a result your measured wind speed may not be a true representation of the perpendicular component.
Add to that the fact that a large component of wind reading is intuition. The wind at the gun may be blowing at 12 to 6 at 8 mph, but what is it doing at the target? What is it doing on the way to the target? If you calibrate your wind observations at the gun (the only wind you can measure) to the true wind speed you should be better able to use environmental queues to gauge wind down range. If you are feeling and observing a 15 mph wind at the gun, but are measuring only 7 mph because because of how you are setting up your kestrel to measure that wind, your reading of the wind down range could be compromised because of the disconnect between what you are seeing and what you are measuring.
What wind is important? Is it the wind at the gun? The wind at the target? The wind 1/2 to the target? Average wind speed? Gust speed? Lulls? Yes. They are all important, but you can only measure it at the gun. Everything else is "reading" and observation. Calibrating your observations to true wind speed "should" allow you to better read the wind you cannot measure.