The curve on your diagram makes me think that you've fitted a normal (Gaussian) curve to the data. By eye, the distribution looks a bit more like log-normal, and so if you're still working on the data, you might want to try that to see. Not that anything you've said or concluded seems wrong, though.

The normal curve in the image is there for operators to visually check if the machine results are normal distributed or not. It's a "stencil", not data.

The software actually does a Lilliefors normality test which returns a big No on this data.