After running Disk Cleanup on the source drive and using it to delete the listed files, and then clearing non-current restore points and shadow copies, I ran the clone again using the same settings.
This time I got much more reasonable results:
Source drive

Clone of source drive

There is still a ~35MiB discrepancy that is potentially worrying, not least because it has the potential to amount to a lot of important documents, but this could possibly be explained by Paradoxon's point on the space that inode tables take up on the file system when it's expanded. I wasn't aware that it was possible for drives and their file systems to have non-file structures that take up data, so this is news to me and I hope it's the case here.
Nonetheless, it's definitely disappointing that a bit-for-bit clone doesn't actually mean a bit-for-bit clone; for me, it tends to remove a lot of the ease of mind of creating drive clones and images instead of manual backups.