Ah, that makes much more sense - and also explains the confusion I had with your emailed answer to my previous question about the format the time was shown in.
It doesn't necessarily explain the average transfer rate error though.
Also, after realising how slow the process was through Opus, I did the remaining 80% through Explorer and it completed in 10 hours (less than Opus did 20%). I've seen mention of copy_buffer_size and copy_nonbufferio_threshold in the Slow Copy Function thread ([url]Slow copy function?]) so will look into that.