The “problem of non-uniqueness” objects to the idea that an empirical quantity is characterized by a unique “true value”: representational structures involving single-valued quantities in fact conflict with the complexity of the systems under examination. The roots of the problem are tied to the model-based nature of measurement. Metrologists answered it in terms of uncertainty, but erred by providing an epistemic answer to a representational problem. One should rather look at it in terms of the validity and the degree of refinement of the models that are involved in a measurement. I thus pro-pose to draw on approximate truth to reconstruct an alternate understanding of “true value”.