Is it possible to use GPS geographic coordinates & ellipsoidal height not orthometric heights in Aerial Triangulation?
In the context of structure-from-motion (SFM) photogrammetry - using ellipsoid heights for aerial triangulation is preferred because the ellipsoid is a mathematical model and this makes computations straightforward. This is especially so if your survey area is large. You can always output your results later in orthometric heights. In my experience, many SFM photogrammetry software gets the aerial triangulation wrong when given orthometric heights especially over a large area.
Capture your source data in ellipsoidal heights. Take/apply as many ground control points as you can - which also must be in ellipsoidal heights. After aerial triangulation, and "model generation", and when you want to output the result (eg DSM) - this is where you apply the geoid undulation layer, which is usually supplied in a grid file.
Of course SFM can only output DSM (which includes heights of man-made structures) and not DTM (which excludes man-made structures). For DTM, you need LIDAR processes, but they are no longer photogrammetry although the LIDAR outputs are often used as inputs for "adjusting" aerial maps.