The Geostationary Orbital Environmental Satellites (GOES) Soft X-ray (SXR) sensors have provided data relating to, inter alia, the time, intensity and duration of solar flares since the 1970s. The GOES SXR Flare List has become the standard reference catalogue for solar flares and is widely used in solar physics research and space weather. We report here that in the current version of the list there are significant differences between the mean duration of flares which occurred before May 1997 and the mean duration of flares thereafter. Our analysis shows that the reported flare timings for the pre-May 1997 data were not based on the same criteria as is currently the case. This finding has serious implications for all those who used flare duration (or fluence, which depends on the chosen start and end times) as part of their analysis of pre-May 1997 solar events, or statistical analyses of large samples of flares, e.g. as part of the assessment of a Solar Energetic Particle forecasting algorithm.