Thank you for your response and for validating that there are quirks in the interpretation of .sf2 files. I was baffled until, on a lark, I tried changing the loop points.
As for the dampening of harpsichord strings, most instruments are regulated so that the dampeners are
just above the plectrum. We don't want much if any perceivable delay between when a key is pressed and when the string is plucked, and the jacks basically hang by their dampeners. So the dampening is pretty quick!
It is, of course, possible to half-release a key and catch it before the dampener engages. I taught my teacher this trick. But is definitely the exception rather than the norm.
Then there's the complication--not accounted for in the samples I have--of a still-vibrating string interacting with the plectrum on the key's release. The norm is a faint and barely detectable buzz. I don't think it would be possible to sample this correctly however. It would require exact foreknowledge of when the key was released and when the plectrum hit the string on its return stroke.
Then again, everything we do with instrument sampling is a compromise, isn't it?