First, because of the manner in which radon concentrations are distributed locally, within a state or county with a specific GM, there will typically be areas that have considerably higher (or lower) GMs. More importantly for members of the public, within any such area, the concentrations found in individual houses vary substantially--consistent with a lognormal distribution--typically including concentrations that are several or occasionally factors of ten higher than the GM for the area.
Secondly, many of the GMs cited from this project are considerably lower than the values widely publicized about indoor radon in specific states or areas, because our analyses are aimed primarily at providing estimates of long-term (specifically, annual-average!) indoor concentrations in the primary living space of homes. In contrast, much of the publicity about indoor radon has been based on results from short-term (usually 2 to 7 day) winter measurements, typically taken in basements where such houses predominate. Because indoor concentrations tend to be higher in winter and in basements, such screening measurements exaggerate the indoor concentration, compared to what occupants are actually exposed to over the long term, which is what is significant from the health point of view. And because of the variability in concentrations, the short-term measurement protocol increases the percentage of homes that appear to exceed any particular annual-average concentration of concern, such as the 4 pCi/l (148 Bq/m^3) EPA remedial action guideline or the 20 pCi/l (740 Bq/m^3) at which occupants receive annual exposures approximately equaling the dose limit for workers in the nuclear industry.
Finally, note that these GMs are always slightly less than the average (i.e., arithmetic mean) concentrations for the areas being considered because indoor concentrations distributions are characteristically nearly lognormal, implying a long tail to high concentrations, which--relative to a symmetric distribution--raises the average more than the GM. Thus, for example, a county GM of 1 pCi/l (37 Bq/m^3) in a state or region where the geometric standard deviations for homes within counties are approximately 2.1 implies an average (i.e., arithmetic mean) that is approximately 1.3 pCi/l (49 Bq/m^3).