Hello Ilari,
do your images contain GPS data? Did you georeference the scene with both GPS coordinates from images and the ground control points? The GPS from images are usually with lower precision than groung control points and if you are using both coordinates for georeferencing there could be a mismatch. You can disable the camera priors before alignment and geo-reference the scene only with using the ground control points or set a higher weight for ground control points’ coordinates.
Also, if it is possible, please post a screenshot of your problem so that we can better understand the problem.