One practical solution is to create two different metadata files, using one for differential analysis (with the unpaired sample removed) and one for the other analyses (with all samples left in).
I would need more details to understand what you mean about “Blocking factor” (see our post guidelines)