Case-control studies on lung cancer and residential radon exposure were conducted in West and East Germany. Odds ratio (OR) estimates from primary analysis are now subject to correction for measurement error in exposure. We apply the regression calibration method adopting a bivariate multiplicative error model of the classical type; that is, we investigate the impact of errors in the exposure of primary interest, radon, and of errors in the most potent confounder, smoking. The OR estimates per 100 Bq/m3 are throughout higher after correcting for errors in radon exposure (e.g., 1.02 and 1.11 for the West and the East German study, respectively, corrected for an error of size 0.4); ignoring the clear but small correlation between radon exposure and smoking of about -0.06 would lead to less conservative corrections (1.10 and 1.13 for West and East, respectively). Accounting for a realistically sized error in the smoking variable additionally increases the OR estimates slightly. Remarkable is the fact that the naive OR estimate of the West study of 0.97 exceeds unity after correcting for errors in radon exposure larger than 0.3. We conclude that correcting for errors in radon exposure has a meaningful impact on OR estimates, that the correlation between radon exposure and the smoking variable affects the correction even if the smoking variable was error-free, and that such an analysis is extremely valuable to grasp an important issue in epidemiology, that is, the dimension of residual confounding due to adjusting for an imprecisely measured smoking variable.