Mark gets it right on the nose, what I refer to as "bracketing". I do not know what these numbers should be.
In the example I mentioned where the validation ride found that a course was .0028% long, this strikes me as too long, for the reason I mentioned. Another performance on another certified course of the same nominal distance that was just a couple of seconds "faster" would in reality be slower if the second course were to be validated at less than .0008 long, for example. The record holder would then lose his/her record to an inferior running performance. This may have actually occurred, for all I know.
Examples:
For a 10 Miler: Race course 1: validates at .0028 long (45 meters); Race course 2: validates at .0008 long (12.88m) difference: 32 meters, or ~ 4 seconds for an open man at closing speed. The shorter course awards a new record to a man who runs 2 seconds slower.
Marathon: Race course #1 validates at .0028 long - 118 meters long; Race course 2 validates at .0005 long - 21 meters long. The difference of 97 meters (106 yds) means the runner on course #2 has several extra seconds to break the record set on course #1.