This original study aims to quantify the human factor uncertainties in radiation doses for Chernobyl cleanup workers that are associated with errors in direct or proxy personal interviews due to poor memory recall a long time after exposure. Two types of doses due to external irradiation during cleanup mission were calculated independently. First, a “reference” dose, that was calculated using the historical description of cleanup activities reported by 47 cleanup workers shortly after the completion of the cleanup mission. Second, a “current” dose that was calculated using information reported by 47 cleanup workers and respective 24 proxies (colleagues) nominated by cleanup workers during a personal interview conducted more recently, as part of this study, i.e., 25–30 years after their cleanup missions. The Jaccard similarity coefficient for reference and current doses was moderate: the arithmetic mean ± standard deviation was 0.29 ± 0.18 (median = 0.31) and 0.23 ± 0.18 (median = 0.22) for the cleanup worker's and proxy's interviews, respectively. The agreement between two doses was better if the cleanup worker was interviewed rather than his proxy: the median ratio of current to reference dose was 1.0 and 0.56 for cleanup workers and proxies, respectively. The present study has shown that human factor uncertainties lead to underestimation or overestimation of the “true” reference dose for most cleanup workers up to 3 times. In turn, the potential impact of these errors on radiation-related risk estimates should be assessed.