okay so this has been driving me insane for like three weeks. doing a mapping project on a construction site, about 14 hectares, using my M3E and running it through Pix4D. i placed 8 GCPs with a Emlid Reach RS2 and im getting sub-centimeter readings on the rover so the coordinates should be solid. but when i process the map the whole thing shifts, like not a lot but enough that it doesnt line up with the existing site drawings and the client noticed.

checked the GCP coordinates like four times, theyre going in correctly. the checkpoints im using as independant verification are showing 4-6cm horizontal error which is way outside what i need. running PPK on the image positions so i thought that would help but its almost like the GCP correction and the PPK are fighting each other somehow? or maybe im just setting something up wrong in the processing pipeline.

base station was set up on a known benchmark both times so i dont think its a datum issue but at this point im not ruling anything out.

The PPK + GCP thing can definitely cause weird behavior if you're not careful about how Pix4D is weighting them. If you have good PPK positions on the images you can actually get worse results by adding GCPs unless you tell the software to treat the image positions as highly accurate and let the GCPs just do a final refinement rather than a full correction. There's a setting in the processing options for image geolocation accuracy ? try bumping that up and re-running. Also worth double checking you're not mixing coordinate systems anywhere, like if your benchmark is in a local grid and your RS2 output defaults to a slightly different realization of the same datum that can cause exactly this kind of systematic shift.

what format are you exporting the GCP file in and are you sure the column order is right? i know it sounds dumb but i once had this exact problem for days and it turned out i had lat/long swapped in the csv. Pix4D doesnt always throw an obvious error it just quietly puts everything in the wrong place.