In capability analysis often only the upper part of the observed distribution is relevant. So, why not add an option to fit just the upper part?
This will make the fit often much better in describing the part that matters, using e.g. a half normal distribution.