Being a compusively data driven personality type I histogrammed all the speed numbers I could confidently read, and the distribution wasn't way off normal, arithmetic mean, median and mode all close to 140; which suggests that the timing data gathering and subsequent speed calculations were decently reliable. (There was a skew, with a tail to higher speed values: could be accounted for by well run in specimens that were that bit more free running, or little run specimens that were more steam tight; other hypotheses are available, we cannot validate without more information!)
If we assume that the distance run - on which the calculation had to be based - was accurate, then Hornby do appear to have pitched the mechanism about right for speed capability.
And quite seriously folks, I have seen business decisions in the many millions of dollars per annum range made on the basis of flakier data than this...