The development of gas separation processes dealing with very low concentration ranges is a rapidly growing domain with key applications such as trace detection, air purification from harmful pollutants, etc. Yet, the design of efficient technologies in this field is hampered by the lack of robust strategies to predict the gas selectivity of optimal adsorbents from simple pure gas adsorption data. Here, the selectivity predicted using different methods, namely Henry's method and the ideal adsorbed solution theory (LAST), are compared with the true selectivity obtained using breakthrough experiments. As a case study, these methods are discussed when applied to Xe/Kr separation in two different process conditions using different adsorbents (an active carbon and two silver-doped adsorbents). Typical data show that Henry's method, in which selectivity is assumed to correspond to the ratio of Henry's constant measured for each gas of the mixture, should be considered with caution as it is very sensitive to the pressure range considered but also the number of points used for affinity assessment. LAST is found to be more accurate provided its applicability to predict gas coadsorption from pure gas adsorption data is first established. However, even when applicable, the case of very low concentrations remains a problem as it leads to very large uncertainties in the selectivity predicted using LAST. We discuss how typical errors in assessing the selectivities using the different methods lead to nonoptimal adsorbent choices for a given separation process. Finally we demonstrate that Ag-loaded zeolite shows xenon capacities and Xe/Kr selectivities that surpass all other materials.