In this answer to an earlier question of mine regarding the predominance of image-type glideslopes (which bounce the lower glideslope beam off the ground rather than transmitting it directly up and out) over end-fire glideslopes (which use a separate antenna installation for the lower beam and don't need to bounce it off the ground), @UnrecognizedFallingObject pointed out something surprising about the differences between the capabilities of the two types of glideslope:
Note also that there are multiple flavors of image-type glideslope antenna (null reference, sideband reference, and capture effect), and some are more tolerant of poor site conditions than others. Furthermore, even though endfire glideslopes can work in situations where an image-type glideslope cannot, they have their own limitations as well. For instance, while well-sited image-type glideslopes can provide the tight tolerances required for a CAT IIIb ILS, end-fire antennas are limited to CAT II operations under ideal conditions, and systems designed for the most adverse sitings are apparently limited to CAT I usage. [Emphasis added.]
Given that both an image-type and an end-fire glideslope transmit the upper glideslope beam in the same manner, and that an end-fire glideslope transmits the lower beam directly up and back along the approach path, while an image-type glideslope has to bounce it off the ground first, it seems that the end-fire glideslope should have the easier time avoiding distortion or interference, as the end-fire glideslope's lower beam
- is not reliant on reflection from the ground, eliminating the potential for beam distortion from irregularities in the ground surface, and
- spends much less time in close proximity to the ground (as it starts moving upwards as soon as it leaves the antenna, whereas an image-type glideslope's lower beam is transmitted downwards, descends to ground level, bounces, and only then starts moving upwards), greatly reducing the potential for interference from terrestrial sources like airport vehicles or local wildlife,
which would, at first glance, indicate that the end-fire glideslope should be more able to provide tight ILS-III tolerances than is the image-type glideslope.
Yet, apparently the opposite is the case (and, frustratingly, the FAA document linked in the abovequoted answer makes no mention of this problem) - why?