Statistical matching aims to combine samples drawn from the same population, where each sample contains information only on some of the variables of interest. The lack of joint observations produces uncertainty about the data-generating model. The paper is devoted to the study of uncertainty in statistical matching for complex sample surveys when a proxy variable is only available in one sample. Such a variable can be used both to verify the conditional independence assumption and to provide a set of plausible estimates of the distribution of variables not jointly observed when such an assumption is not satisfied. Finally, a simulation study is performed and an application to integrate objective and subjective well-being measures is provided.