Yes, exactly, I was quite surprised to find out that RA has the data, and yet is not utilized. I was equally surprised how old actually GW data is, assuming that it was available since it was obtained which was 2011, which was before Assetto Corsa. And people still go with majorly off conversions.
I think only minority could tell if track is on scan data, or if it is well done but without it if it wouldn't be stated so. I guarantee Kunos Zandwoort would have passed as laserscanned, just as example. Furthermore, I am not a specialist about Road Atlanta, so it seems good to me, I was same about Goodwood, but immediately after starting paying attention to real thing more closely I found that it is incredibly off, and later LiDAR data proved it.
And it is so much harder and time consuming to do something without scan, and yet that dark cloud that it is not scanned (but could be), thus potentially off somewhere is always floating around.
It is definitely not worthy to model any track without LiDAR if there is a slight possibility to get the data.
IMO there are three main kinds of accuracy error in tracks. Localised errors where at particular point track is majorly off, relative errors where track is good in particular segments, but those segments in relation with each other are off, and then detail errors like cambers, width, smaller bumps, roughness, kerbs, maybe even surface friction levels - takes months without scan data, even with a help of dozen real life racers. With lidar only small scale stuff is left for free interpretation. And with ground level scan it would be basically only processing of data and art.