Dose conversion coefficients attempt to harmonize the material-, location-, and exposure-dependent results from retrospective dosemeters. The issues and uncertainties arising from dose conversion are explored within the framework of an interlaboratory comparison exercise in which mobile phones were positioned around anthropomorphic phantoms and exposed to non-uniform photon fields, with the glass and resistors they contain employed as fortuitous dosemeters. The difficulties of adopting pre-calculated tables of generic conversion coefficients are evaluated first, and then compared against those arising through the use of bespoke data derived by Monte Carlo modelling, and also against not converting the doses measured by the phones. It is seen that the different subjective choices that users might make when selecting ‘optimal’ generic data can lead to a significant source of uncertainty (up to around 70%), though may be improved (to around 30%) by appropriate quality controls. Use of generic coefficients typically led to over-estimates of the organ doses: an average discrepancy of ca. a factor of 2 was found, but this is still better than the factor of around 3 observed when no conversion coefficients were applied. Use of bespoke conversion factors led to the best estimates of organ doses, although they still over-estimated by approximately 1.5 on average, and an uncertainty of around 20% was associated with generating their values. Overall, applying bespoke conversion data improves but does not guarantee correct dose categorization of individuals, with the inconsistences in the measured results found generally to be the limiting factor in obtaining accurate dose assessments.