Hi Akin
the easiest way is to get more images overlapping the “empty space” of both COMPONENTS if it is possible. If you cannot get more images, then you need to use control points (CPs) on both datasets, for each CP you need at least 3 projections per component, so 6 projections as the minimum, more projections mean more help to get it together.
Take a look at how good is to place and spread CPs in images: https://www.youtube.com/watch?v=d8naLEtLqDY