Firstly, you introduced this Q in the thread this came from. I have ‘just’ dragged @sciguy 's excellent post as relevant here.
Idk (be gentle, a long time since I was examined in this area) but the measurements between tx element (on fork blade) and rx element (actually both) are achieved with Ultra-wideband (UWB) technology (frequency 3-10 Gigs). If the antenna gain arrays are narrow angle (but both forwards and back) they’ll ‘lose’ the lead rider round a corner rather than ‘see’ a shorter chord rather than the arc ridden.
On the chord/angle issue, this is fairly irrelevant on a 90o turn. Coming into the turn which will have a cycling line radius of curvature of (say) 6m (really tight), the difference between the chord and the 90o arc (quarter circle) is <1m.
But, and idk, I hypothesise that in 180o turns, there is the possibility (and this may be a ‘ghost’ yoy-yos source) that the two riders’ RRs generate a <12m start time and finish time. Some of the T100 shared (and IM athletes individual) data has multiple yo-yos which added together imply fractions of a second for each yo-yo. more likely this is just picking up concertinaring at 180s.
I ‘liked’ @marquette42 's observation, including the identification of the drafting=cheating zealots cohort (none on here?).