Yes, exactly. Using this option BLAST will only report hits with 100% coverage having sufficient expect values. Assuming they exist, which they won't.
BLAST is a local alignment tool, it is only designed to find local alignments. BLAST will only concatenate HSPs if the score of the new segment is significantly (or just greater, depending on the specific version) greater than the score of the current segment.
Basic Local Alignment Search Tool: Protein BLAST (BLASTp)
Gaps and mismatches have penalties associated with them. So, if a new segment requires too many gaps and/or mismatches, the score of that new segment will be less than or not significantly larger than the current (shorter) best hit.
The shorter hit has a higher score and represents the local optimum and is what is reported.
Setting qcov_hsp_perc doesn't change how BLAST searches for things, so if OP isn't finding 100% coverage hits already, setting this won't change anything. Outside of a few specific situations, I doubt there's a case where BLAST would extend to cover 100% of the query and score it lower than a shorter hit.
Basically, OP will see the same results as before, with hits having less than 100% coverage being omitted from the results.
Thanks for these precisions.
I understand your point now.