This is wrong, terrain heights only exist in a range of several thousand. When you multiply the X and Y coordinates by 1,000,000 each, the incorrectness of the Z orientation would become completely negligible.
And that's if what you're saying were true, but double check the function... the unit is looking at itself with an offset, meaning it already has its own Z height factored in.
It works fine when I use it in my physics engine.
Bah, right of course
I don't know what I was thinking ...