Test charts are a tool commonly used for the evaluation of colour reproduction system accuracy. They may vary in size and choice of values tested. Within statistical quality control, the importance of random sampling is very often stipulated. However, in the context of the evaluation of colour reproduction system accuracy, samples are very small with respect to the number of possible device inputs. It is therefore very common that practically usable test charts of sizes of a few hundred patches underestimate the maximum error. Including a larger number of critical values such as those at the gamut boundary increases the probability of estimating the maximum error more accurately. However, the sample is no more random and the inclusion of a larger number of critical values inevitably increases the central tendency measures. The aim of this research is to improve the maximum error estimate, while retaining the central tendency estimates within the range of statistically insignificant differences. This is achieved by adding critical values to the chart of sufficient number of patches.