The prior pose is the image’s pose taken from the EXIF data. It can be also modified by flight log.
For drone data this is not very precise (unless it is a RTK drone), and therefore the georeferencing can be moved or not scaled properly.
In that case this prior pose could be used just for the first alignment and then the GCPs are used for proper georeferencing and alignment (as GCPs work also as control points). In this case the prior pose could be different to an actual pose and therefore for second alignment it is set to Unknown. Then only information from GCPs are used to georeference and scale the alignment properly.
May I ask why you didn’t turn off the prior poses as it was recommended to you before (in your previous questions from July)?