The variations seen in the tweeter curve, with the exception of the bump from around 12k-18k, are almost exclusively due to baffle and woofer diffraction. I found that even the woofer will cause some diffraction. If a smooth cover is placed over the woofer cone opening, some levelling occurs. This is small in comparison to the baffle edges.
Note the undulations in the on-axis curve. Then look at each of the 15° measurements and note the location and amount of variation. Then look at the average. Since the perceived sound is a combination of the direct and reflected sound, the off-axis has a significant influence. The effect depends upon several factors, amoung them the proximity of reflecting surfaces such as side walls and the amount of damping on those reflecting surfaces.
To my ears, the most significant impact seems to be in imaging. If the average response is smooth, then the imaging depth is much better. It's really quite surprising.