The unit of measurement is already known, by agreement an Imperial horsepower is just shy of 746 watts. That agreed conversion factor has never been in question

The unit of measurement is already known, by agreement an Imperial horsepower is just shy of 746 watts. That agreed conversion factor has never been in question as we've never cared what the actual power output of a nominal horse is under nominal (defined) load conditions. My point is the actual power of a real horse in agreed horsepower units cannot be accurately determined if the conversion of 'pulling' force isn't accurate, so the only issue is calibration. Even in the metric-phobic US the rest of the dynamometer would likely be designed around SI units simply because that's how most engineering measurement is done nowadays. That means all calculations would be done in SI and converted into legacy hp units. Fortunately, Imperial measurements in science/engineering has all but died so the interest here is only academic. Another similar anachronistic measurement is the illumination unit candlepower, again a conversion factor to SI was agreed but in this instance (unlike hp) there was a defined physical standard—the international candle (or it was so some 80 years ago until the standard was updated). (I say 'fortunately' above as I had to learn both Imperial and metric and I'm damn glad Imperial has all but gone. Dealing with foot-poundals, PSI, etc. is for the birds.)