I have reproduced your BLAST analysis using the NCBI web service, using your list of 10 positive strand ORF's for CoV_11241010.
I'm observing a classical number of hits (not 2 or 3 hits only as we observed in your results at the end of Class 3), for instance for your ORF4:21476:25201, I get 500 BLAST hits in RefSeq (ie the limit I chose in the BLAST parameters), with E-values ranging from 0.0 to 2e-08.
This confirms my suspicion that the results you showed me this morning resulted from a NCBI BLAST bug, could it be due to the fact that the computation was carrried out on a tablet rather than a laptop computer (maybe the tablet doesn't have sufficient hardware resources to deal with such complex web pages) ?