DM-34693 fixed most of our missing solar system object matches (except for one object, 2007 TC479, which appears to be genuinely undetected). However, the predicted positions for these objects are still systematically off by approximately one arcsecond, which is usually more than the predicted position error.
This is shown in the attached DS9 screenshots, where the circle represents the position at the exposure midpoint and its uncertainty (unfortunately, I have no data on the shape or orientation of the error ellipse) and arrow represents the proper motion expected over the duration of the exposure. North is up in all images. 2003 SS69 is from ap_verify_ci_cosmos_pdr2; all others are from ap_verify_ci_hits2015.
The detected object is always "behind" the proper motion vectors, but I don't know if this is significant, since all these objects were observed in the same part of the sky and have similar proper motions.