For starters - we typically are not determining velocity for a statistically significant number of shots. The ES/SD from a 5-shot group velocity determination simply doesn't carry much statistical weight. Unfortunately, most of us don't want to burn out our barrels doing load development, so we are usually forced to work with sub-optimal velocity sample sizes.
Second - low ES/SD may not mean much if your load is "out of the optimal window". It has been clearly demonstrated that the effect known as "positive compensation" occurs, and that it can be responsible for lesser vertical dispersion at long distance than would be predicted solely from ES/SD values. The short explanation of positive compensation is that bullets with slightly higher velocity leave the barrel slightly earlier in the harmonic cycle, when the muzzle is earlier in its upward travel and therefore isn't pointing upward quite as much, so they exit the bore with a lower launch angle. Bullets with slightly slower velocity leave the muzzle a tick later when it is pointing at a higher launch angle. Because the slower bullets exit the bore with a higher launch angle, the arrive on target at a similar vertical height as the slightly faster bullets that exited the bore at a slightly lower launch angle.
The keys for positive compensation are that A) it doesn't just happen...it has to be specifically tuned in to the rifle barrel harmonics via proper load development, and B) it apparently can occur over a velocity range (ES) of as much as 20-25 fps or so.
There are a couple take-home points from this. The first is that as long as the load is properly tuned, it is possible to experience the benefits of positive compensation without being forced to maintain extreme spreads under 10 fps. The second point is that when a load with an ES in the ~15-20 fps range shoots tiny groups and is likely benefitting from positive compensation, but a load with a much lower ES of 5-10 fps shoots much larger groups, it is unlikely that ES has anything to do with why the load with the lower ES/SD shot larger groups. In fact, positive compensation in a properly tuned load should occur the same whether the ES is 15-20 fps, or 5-10 fps. So the likelihood would be that either the load with the lower ES/SD that shoots the larger groups is either not properly tuned with respect to barrel harmonics and positive compensation, AND/OR the sample size of the groups used to generate both ES/SD values are simply not optimal statistically. In other words, would the ES/SD values of the load that shot larger groups really be any different from the the load that shot tighter groups if something like 25 or 50 shots had been fired?
The final consideration is that different types of competition shooting events may require somewhat different reloading approaches. For example, F-Class shooters typically fire 20-25+ shot strings in a single relay, and will shoot 3 relays during the course of a single day's match shooting (i.e. early morning to mid-afternoon). It is not uncommon for the temperature to change dramatically during that time period. Trying to excel in F-Class competition at 1000 yd with a load that has shown 20-25 fps ES (or more) for 5-shot groups during routine load development will be an exercise in frustration. In fact, such a load will likely exhibit an extreme spread of more like 30+ fps over a typical 25-shot string. The loads we develop for F-Class don't typically possess quite the precision necessary for top 1000 yd Benchrest competitors, but the velocity needs to remain relatively stable over the longs strings of fire, throughout the match/day. So F-Class shooters generally tend to pay a little more attention to velocity and ES/SD data. It's pretty simple really, your reloading approach needs to work well for the chosen shooting discipline, and there is usually some latitude within any approach that can work successfully.
