About to head out but in short:
1) Add your images and set the correct coordinate system in the reference settings and set marker accuracy to 0.02 m.
2) Align images with High accuracy, generic pair preselection 40k key point limit and 0 tie point limit.
3) Load GCP's in the reference panel. GCP.txt should be formatted as label/X/Y/Z columns or any variation on this.
4) In the sparse point cloud, use the selection tool in the area where you believe a GCP is located (takes some guessing and investigating). Then right click and press Filter photo's by selected points.
5) Then in the photo pane just click on the newly selected images and locate your GCP and then right click and place the appropriate marker. Do this twice for the GCP and then right click on the GCP in the reference panel and select Filter photos by marker.
6) Then continue to mark all the images for that particular GCP. When done repeat step 4 and 5 until you've fully marked at least 3 GCP's.
7) Use the button 'Optimize Cameras' in the reference panel, leave all boxes at default (K4 is for fish eye lenses).
8 ) right click on any still unmarked GCP in the reference pane and right click and select Filter photos by marker and mark *all* the images for *all* the GCP's.
9) Use the Optimize Cameras again, leave all at default.
Now you should end up with a sparse cloud that's fully georeferenced and it shouldn't show any bowl effect any more.
10) Build dense cloud with *low* density and aggresive depth filtering.
11) Use ground classify points; I usually use the parameters 18/0.3/100.
12) Build mesh; enable interpolation and *ONLY* use the *GROUND* points.
13) Build Orthomosaic in the resolution you desire.
14) Double click on Orthomosaic in the workspace panel to edit seem lines.
15) Update Orthomosaic once you're finished with seem line editing.
16) Export Orthomosaic in the resolution (X&Y scale) you used above
17) Done, artifact free geometrically calibrated Orthomosaic