Implementing positioning systems for indoor environments is notoriously difficult. The vast array of interacting parameters such as furniture, room shape and materials mean that the simulation of such an environment is of limited value. We present in this paper results from an empirical investigation as to the attainable accuracy of a Radio Frequency (RF) positioning system based on received signal strength (RSS). Methods to improve confidence, such as frequency averaging and bi-directional ranging, are explored. Finally we present and evaluate a novel method for calibrating the channel propagation exponent with no prior knowledge of the sensor network layout or room shape.