Sorry about that, folks -- I meant to have this expire on the 30th, not the 28th, to space things out a bit with the other puzzles. The expiry date has been updated.
In this very small puzzle (I have never satisfied a core filter with such a peptide) won't the bridges be exposed and easily attacked by reducing agents in the blood and CSF? I suppose I can *try* to stuff the bridges in the middle of the core, but again, I probably don't have a core, at least, how the filter defines it.
The blood and spinal fluid are oxidizing environments, so we don't need to worry about the disulfide bonds being reduced. Any reduced cysteine pairs would quickly oxidize.
It seems unlikely that loops will fold in a tiny knot the same way as globular proteins. Perhaps if you are going to make many different peptides in the laboratory, you could compare the effectiveness of the fragment filter on this puzzle. It would be great to turn this off if it is not helping.
You make a good point that heuristics like the Fragment Filter probably don't apply equally to both large globular proteins and small peptides. Unfortunately, it may be the case that the Fragment Filter is more important here.
In fact, someone in the Baker Lab has recently done an experiment similar to the one you propose. They found that for small disulfide-bonded proteins, fragment quality is actually a powerful predictor of protein stability in the lab.
One possibility is that "bad" (i.e. rarely observed) fragments perhaps represent subtly strained conformations of the backbone. A large protein may be able to compensate for this strain by making a lot of favorable interactions elsewhere. A small protein would simply adopt another conformation with less strained backbone.