We have Litz's methods published, do you have access to that of Hornady? Are they using the same methodology? I've never seen the actual testing method used by Hornady.
I've only seen their dissection of confidence, which is not terribly advanced statistics, but does reveal to anyone familiar with these methods as to why we really can't take faith from 3 shots - especially when we KNOW a great SD from our 10-20 round strings is 4-7fps, and each data point on the velocity curve has to be positioned in a specific 2-3fps band to produce either a ramp or a flat spot... We KNOW our SNR is very low, in that analysis.
In general, and maybe it's a cynical view, I'm actually prone to expect most of us are always stuck using philosophy more close to CCD rather than FFD anyway, even when folks attempt FFD - simply because we don't have control over too many aspects of influence (roundcount on the barrel, roundount since cleaning, differences in individual barrels, environmental conditions, etc). But I really think it's much simpler than that - this acknowledgement of the population noise vs. the variable dependence of the results is where most of us are simply lying to ourselves, and there's just no way around it. If we move any data point by 5fps on any of our flatspots (let alone floating ALL of the data points around by 5 fps, in either direction), faith in the flat spot dissolves really, really quickly. How many 20 shot groups have any of us shot which had an ES less than 10? Hell, it's rare enough to see 20shot SD's less than 5 (+/-5fps SD = 10), let alone the reality that ~1/3 of our expected readings should be OUTSIDE of +/-1SD. When our SD is say, 5fps - Standard Deviation, meaning the standard value which each value deviates from the average for a given sample set - how are we supposed to trust that any data point is really within the 1-3fps we need it to truly represent to make a flat spot survive?
Another good example a guy offered me once when he was trying to convince me the Satterlee Curve didn't survive - what happens if you shoot 3 strings (which are usually shot round robin, one of each charge), and you see those "recurring flat spots," but then you randomly remix the shots of each charge into synthetic strings? We shoot round robin to randomize the results, so random velocities for each charge weight SHOULD be able to be intermixed together and repeat the flat spot, right? So why if we randomize the speeds for each shot within a charge weight, why don't the flat spots persist?