Hello River123. I've been kicking the PhotoScan Pro tires w/ a trial license for a week & a half & learned a few things about camera position but am still fuzzy on some of the items. To answer your question, based on what I've learned, it appears that it's best to load the camera position info before performing any photo alignment. Have you seen the PhotoScan tutorial
http://www.agisoft.com/pdf/PS_1.3%20-Tutorial%20(BL)%20-%20Orthophoto,%20DEM%20(GCPs).pdf ? This helps some but makes assumptions as to how familiar you already are w/ the software.
What I'm struggling w/ is how PhotoScan interprets or describes the camera position parameters: lat, lon, altitude, yaw, pitch, roll. I'm taking them literally given that I'm an aerial reconn camera guy. Thus in my world these parameters are specific to the platform attitude. However, in my camera the lens assembly is mounted in a 2-axis gimbal. Thus the camera line of sight also has an attitude of pitch & roll. So, should the camera position parameters be a combo of the platform + camera attitudes? Likewise w/ platform vs. image position. The platform yaw, pitch & roll are based on nadir, yet the lens may be projecting its line of sight to the ground at some oblique angle that is a function of camera lens pitch & roll. Therefore should the camera position lat, lon & altitude be from the image or the platform? See my conundrum?
I've adjusted these parameters in a multitude of combos: platform location values only; platform + camera attitude; ....
None of the combos seems to be just the right combo for a clean alignment of a pair of overlapping images & thus generation of a dense point cloud (FYI - my end game is to generate a DEM from overlapping images).
I've reviewed the above in the PhotoScan Pro manual as well has posted these questions to PS support. Neither source has been very helpful. I'm hoping this Forum site might? Paul