Yes, I've also noticed that source of error. However, I don't know how you get around that. Depending on the curvature of your sit bones , the tilt of your pelvis, and the shape of the seat, the exact contact point on the seat and your body can shift a bit regardless of the length measurement. For me, I read up on my particular seat and it's said to tend to set you back about 1 cm more than the measured value, compared to other "average" seats.

I'm not sure where the sit bones contact is the widest part. It depends what you mean by sitbones, and what you mean by contact. What most people think of as sit bones (the protrusions you sit on on a bench) don't carry weight in my normal position, and the part of the pelvic bone that does carry weight certainly isn't at or near the widest part. However, the "sit bones" probably are approximately over the widest part (maybe slightly in front), and if I do sit up straight to rotate back onto my sit bones, then they do contact approximately there. I'm not sure this approximation does much better than the nose though. Maybe a little.

I'm afraid some errors do in fact just have to be fine tuned in the fine tuning and you better size the bike with a little wiggle room in all directions, like with a predicted stem of 110 to give a little room for play either way.

Still, I think nose to handlebar is actually the most meaningful number a calculator can probably produce. It's probably slightly more meaningful as a 2-d distance rather than a horizontal, but unless you have 20cm of bar drop the difference isn't huge, and I'm assuming the same arm angles are optimal regardless of how tilted you are, which might not even be true. In the extreme case of doing a hand stand, I don't think arms at 90 degrees works very well. So it might make sense to stretch slightly anyway if dropping down that far? That's beyond my expertise.

While a calculator can likely tell you about how stretched out to be, I don't see how it can tell you how setback you should be, and that directly impacts frame reach and/or stem length. I also don't see how it can tell you how leaned forward you should be.