I see what Pete's and Matthew's concern is, but I think Neville has happened upon two distances where different methods of riding produce the same statistical wobble.
If a rider rides along a 300m course, his counts per km differed by such a miniscule amount from a 25m course which was coasted along, that a series of "coasted" cal rides would yield virtually identical "constants". In his rides, his 300m rides yielded a 479.74 clicks/km, vs 479.75 clicks/km in his 25m rides.
This small test would indicate that riding a 25m course very carefully (not varying the route by more than 4") yields the same wobble percentage as riding a 300m course normally. If this can be replicated by at least 5 other riders, then I think this approach should be examined further. Mathematically it makes sense, and logically it makes sense.