On the appropriate and inappropriate uses of probability distributions in climate projections and some alternatives