How about the comparative number of files backed up, vs. the total bytes? Lots of smaller files means a slower backup and compare than fewer larger files.
If there's a significant difference in the processor speeds between "AG-04" and "DD-13", that will also affect the Retrospect "client" backup speeds.
This post in a 2017 thread, especially the first two paragraphs (the P.S. mostly rules out speed of the "backup server" for "client" files), covers the first part of experiments I ended up doing that have a bearing on this. This post in that same thread, and the one following, covers the rest. This post near the end of the same thread discusses my hypothesis on slowness of "client" backups.
P.S.: This ArsTechnica Mac forum post says "My 2012 Era iMac with 1Tb Fusion Drive is starting to really slow down. It seems like when the drive is nearly full, the physical drive doesn't like to spin up anymore. ugh."