Hi Ryan,
26 GCPs that’s a lot!
We do usually recommend between 5 and 10 GCPs as adding more does not improve significantly accuracy.
Step-1-Before-Starting-a-Project-4-Getting-GCPs-on-the-field-or-through-other-sources-optional-but-recommended
Sometimes due to the nature of the images, some area are difficult to reconstruct as there is few tie points.
For example, this can happen in very homogeneous area such as water, ice, desert and forest.
Having a look on your project, it seems that the GCPs showing this shift is located in an area with few ATPs.
Here we can see that this area has very little matching which even if every thing is good will lead to a lack of accuracy:
Personally, I would set some of the GCPs as Check Points in order to see how to accurate is the geolocation. (No need to reprocess, Reoptimization is enough and then generate the quality report to check the results).
Difference-between-a-ground-control-point-and-a-checkpoint
Using-checkpoints
Then, depend on your processing options, accuracy could eventually be improved.
I would suggested you to use to following if it was not already done:
- Set the image scale on 1/2 improves the matching
- “All Prior” for the internal parameter optimization
- Geometrically Verified matching
Hope this will help you to improve your project accuracy and if it does not, could please provide us your quality report?
Best