I thought a bit about the shadow zone shape from the 'other perspective'.
And I came to this:
First situation I consider 'normal'. Listener is on the surface (the black dot) and he is interested about transition loss for different points in vertical plane. This is the situation I measured and described in 'measurement thread'.
Bright means normal signal, middle blue means somewhat worse signal, dark blue means shadow zone.
Now speculation begins. If you reverse the logic, and put listener into the shadow zone (ie. under the layer), and you test to which sources he would appear in the shadow zone, his 'TL map' would look like this:
And if you revert this again, you must get this results to listener at any depth above the layer:
It would need some more testing but I guess it should look like this.