Single numbers are simply, surely obviously, not appropriate results for representing many future uncertainties. This has been well explained by Simon Carne (2004) and by Zyen (2012) and is especially the case when the measure provided is never fully specified. Is it the mean, the median, the mode or some unspecified percentile?
We should be looking at multi-dimensional results with confidence intervals, for which the deterministic approach fails. Our objective should be avoiding the “fog of certainty” (actuarial noise). On its own, that liquidity cannot be detected in advance by a discounting process should be a compelling argument for ditching discounting.