Two UK Schmidt plates of comet Hale-Bopp dust tail taken in 1996 May are analyzed by means of the inverse dust tail model. The dust tail fits are the only available tools providing estimates of the ejection velocity, the dust-loss rate, and the size distribution of the dust grains ejected during years preceding the comet discovery. These quantities describe the comet dust environment driven by CO sublimation between 1993 and 1996, when the comet approached the Sun from 13 to 4 AU. The outputs of the model are consistent with the available coma photometry, quantified by the Afρ quantity. The dust mass loss rate increases from 500 to 8000 kg s-1, these values being inversely proportional to the dust albedo, assumed here to be 10%. Therefore, the mass ratio between icy grains and CO results is at least 5. Higher values of the dust-to-gas ratio are probable, because the model infers the dust-loss rate over a limited size range, up to 1 mm sized grains, and because the power-law index of the differential size distribution ranges between -3.5 and -4.0, so that most of the dust mass was ejected in the largest boulders that Hale-Bopp was able to eject. The dust ejection velocity close to the observations, between 7 and 4 AU, was close to 100 m s-1 for grains 10 μm in size, much higher than that predicted by R. F. Probstein's theory, thus confirming previous results of Neck-Line photometry. This result is an indicator of CO superheating with respect to a free sublimating CO ice, in agreement with the high observed CO velocity. The fundamental result of the paper is that such a high dust velocity remained constant between 13 and 4 AU, thus providing a strong constraint to all models of the CO-driven activity of the comet during its approach to the Sun: CO superheating must have been active since 13 AU from the Sun. It might be provided by the abundant dust itself, or by seasonal effects heating the subsurface layers, as was suggested for comet 29P/Schwassmann-Wachmann 1. Another similarity between the two comets is provided by the power-law index of the time-averaged size distributions: -3.6 ± 0.1 for C/1995O1 and -3.3 ± 0.3 for 29P/SW1. However, other characteristics of the dust environments are very different, so that, in general, it is impossible to distinguish a CO-driven comet from a typical water-driven one.