I think both techniques are required.
1) If you swing a yagi looking for a peak you are likely to get a reflection from a water tower or building.
2) Calculating the expected direction informs you that you are "on the mark" as far as the peak you find. If you have a scanne like the BCD536HP, the bar graph is rather tiny and not intuitive.
If anyone knows different way to skew the 6 dB points accurately with that scanner I would love to know. I guess I could insert a 3, 6 and 9 dB pad and figure out the actual bar that responds.
There will be cases, for those inside the simulcast footprint where the F/B or a side null will be be useful to help reject an interfering site. In that case the nose of the yagi might be off center with the direction. In other words trading 3 dB signal for 10 dB interference rejection.