Reinforced concrete (RC) structures in harbors are designed for long life (50–70 years) and are subjected to chloride-induced corrosion. Diagnosis of these structures usually relies on inspections which consist of ponding salt in the concrete cover with semi-destructive testing (SDT). Uncertainties in inspection result in bad diagnosis, and consequently in bad decisions. This study aims at developing a methodology for quantifying the uncertainty of measurements taken on-site where the random properties of concrete play a role. This assessment was performed through multiple measurements by three operators from two laboratories on the same cores extracted from a 27-year-old existing bridge located in Ireland. A total of 566 measurements are available. The effect of the operator, laboratory, and protocols are highlighted and then modeled. It is shown that the error of assessment is a function of the chloride content and follows a student distribution. The standard deviation of this error reached approximately 15% of the chloride content. This error is further propagated through the Fick law parameters for measuring its effect on the diagnosis: a 17% change of the probability of corrosion initiation is shown. This is the first study of this type with in situ inspections and the work is partly transferable to other coastal or offshore structures.