Hello Ppant,
You can define the coordinate system (i.e. scale and orientation) by placing a few coded targets around the object of interest and having their coordinates input to the Reference pane.
For example, print 4-8 targets on the single A2 or A3 sheet, so that you could easily measure their coordinates, assuming that one of the targets has (0, 0, 0) coordinate. Then put this sheet under the object of interest before taking images. After all the images are collected you can run Detect Markers command, input the coordinates of the detecte markers (or load from the text tile) and apply this information to perform the model referencing.
I recommend to use anything that could help to scale the model properly (at least a pair of markers with known distance between them). In such case you would need to apply the model orientation either manually (using Rotate Object instrument) or with custom script that will do that automatically after manually placing three markers (even in model view) on the flat surface.
If you are planning to follow the latter approach, I can assist you with such script.
Note that in any case it is recommended to uncheck all cameras in the Reference pane and switch the coordinate system to Local Coordinates for the chunk, as in the small scales geotagged images are not helpful.