The projected discovery and exclusion capabilities of particle physics and astrophysics/cosmology experiments are often quantified using the median expected significance. We argue that this criterion leads to flawed results, which, for example, can counterintuitively project lessened sensitivities if the experiment takes more data or reduces its background. We advocate for standard use of the “exact Asimov” criterion detailed in this talk. Various counterintuitive flaws associated with frequentist and modified frequentist statistical measures of significance for multichannel counting experiments are discussed in a general context and illustrated with examples. We argue in favor of conservative Bayesian-motivated statistical measures, and as an application, we employ these measures to obtain the current lower limits on proton partial lifetime at various confidence levels, based on Super-Kamiokande’s data, generalizing the published 90% confidence-level limits. Finally, we present projections for exclusion and discovery reaches for proton partial lifetimes in p→ν K+ and p→e+ π0 decay channels at the Hyper-Kamiokande, DUNE, and JUNO experiments.