In reality, this PhotoField is probably our best built and most accurate one yet! I can certainly understand the confusion above, however.
I'll explain.
PhotoFields are all located in a "world" that is a perfect flat plane. You never see that because visually it's covered up by the panorama. But the underlying terrain is entirely flat in all directions.
"Level" ground in the real world is anything but. There are pretty much always at least minor changes in elevation over larger distances. The amount of variation often ends up being quite surprising compared to what our eyes perceive. And it turns out that even small changes end up wreaking havoc with panoramic images. Also, aside from whatever elevation changes and unevenness may be present in the scene, the panoramic image itself is likely to have some bowl distortion, even if it is well made (and possibly other issues if it is not). When the model gets far enough away, it will either seem to penetrate the ground or be floating above it even though it is taxiing happily, or things will appear too large for their actual distance. All kinds of fun things can happen!
In cases where the terrain elevation changes much at all, and/or when there are grass berms or other similar features, we compensate using invisible 3D objects to match that shape. Then the RC model generally interacts with the world as it should. It's quite similar to how we use invisible 3D objects matching the shape of other objects in the scene like tables or buildings so that you can collide with or even fly behind and under them.
In the past we have very carefully constructed those kinds of terrain objects by hand. It takes a lot of work. AMA Flying Site 3 is the first to use a new, vastly superior technique. We used a detailed drone scan of the site to capture every detail and then constructed 3D terrain objects from that data. It allows us to represent the site with far greater fidelity and nuance. There are numerous advantages, such as the ability to better account for the kinds of elevation changes I mentioned above and to keep things looking correct farther away from the camera.
(We have an unreleased test PhotoField of our local flying field created using this technology, and the way it reproduces the "bump" an airplane makes when transitioning from the slanted taxiway onto the more level runway is uncanny. It really tricks the brain into thinking you're there!)
The only problem is the one you encountered: when the aircraft is resting on an object that represents "terrain", the altitude displays do not take that into account. This results in reported altitude values that are quite inexplicable until you understand what's going on. (Incidentally, you can see the same issue but to a much lesser degree at LIARS Field, where the earlier, more rudimentary version of this technique had to be employed.)
I have filed a case in our bug-tracking system to look into the best way to compensate for that. Thanks for bringing it to our attention.