Indeed, as mentioned by @Adam, we could provide better assistance if you granted public access to your quality report.
In the meantime: on one hand, the Error[s] X, Y, and Z for the check points seem to have higher values than expected, especially since they are higher than 2-3*GSD (as mentioned in the post title). On the other hand, the overall projection error values for all 3D tie points (ground control points GCP and check points) are pretty low. This means the 3D tie points were well marked in the images and that the project was well fitted to the ground control points.
Note for check point 012: you could check out how you marked it on the images.
There could be a number of reasons why the error values for check points is higher than expected, such as:
the location of the check points in the project. The error can be higher if the check points are located on the edge of the project
the location of the check points with respect to the ground control points. The further the check points are from ground control points, the lower the absolute accuracy of the position of the check points in the project.
the flight plan: mainly flight height and overlap
the camera optimization
However, it is hard for us to give you a more accurate response without the quality report. Could you also give us more information about your flight plan?
Additionally, what you should also keep in mind, is that the 3D tie points (ground control/check points) are weighted differently according to the accuracy you indicated. As a consequence, make sure you neither overestimate nor underestimate the accuracy of your ground control/check points.
Looking forward to hearing from you!