the whole 'real human performance requires imperfection' should not be at the level of the recorded source. it is really a function of timing and idiosyncratic imperfection of the pitch, etc, during performance, by particulars, in context. I would never want bad attack per se, bad playing, bad intonation. we are working towards a bit of the ideal in virtual performance. I don't want to have a poor start. It's not very well thought-through, that argument.
I'm good at sequencing because I have studied in depth what happens when people play instruments. it's a very subtle thing really.
I don't know how good Adagio is until I got busy using it. Whoever made the articulation demos isn't good at sequencing string parts, or it's crap. I won't risk money operating under the assumption, 'sure, the lib. has to be better than that'. If the person sequencing built the library chances are good they aren't really a strings person.
I wanted it to be good so there's an option. It's your money, though.