I'd imagine the biggest problem is that radar observations =/= actual wind speed obs, so there's plenty of uncertainty as to how, exactly, that sort of thing should be handled. Radar data may be of varying quality and usefulness from situation to situation, so what sort of rules do you use to determine what's acceptable evidence and what isn't? It probably won't be at or near ground level, so how do you translate data from, say, 50 or 100 or 200m AGL? At what point is it just too far above ground level to be useful as a representation of a tornado's intensity near the surface? Does it have to be a sustained velocity? How long? Is a 225 mph velocity for 3 seconds at 150m AGL enough to assign an EF5 rating?