I am not a member of the Data Lab team and was not aware of this help mechanism. As a publisher in the DL I assume this would eventually come to me. This behavior came as a surprise to me and I first verified it in my master copy. After wracking my brains about this I realized the CCD field is spurious for any dataset that begins with MOV. I’ve used various methods as I’ve tried to exploit the DECam archive for serendipitous asteroid detections. The MOV class of datasets was done by making single images of the DECam field (from the archived stacks I think) and then doing the difference detection on that. In this case there is no info about the CCD but the part that converts detections into database records ends up with something that is meaningless. This is also why the x/y fields are generally large numbers (i.e. not within 2048x4096).
In principle you should be able to do the inverse of what you mention: given an ra and dec and observation invert to CCD coordinates. However, I know that would not be generally easy.
I can try a help you with your goals if you want since the number from DES data is relatively small. Do you want cutouts or something like that? I could probably also create something that would do the inverse mapping so you could get CCD + coordinates.
Thanks for pointing this problem out.