The guys who really value accuracy are over at
www.gps-speedsurfing.com/default.aspx -- they haven't ratified KA72 yet (to the best of my knowledge).
However KA72 gives very similar results to Realspeed, which is ratified. Dylan has worked very hard to get the algorithms working right, and I'm confident that the results that KA72 gives are correct.
The difference between the programs is normally a lot less than 5%, so I wouldn't worry about it too much.
The important thing is to always use the same program for your calculations. If that's one that always gives you higher numbers, then that's fine. I like to draw the analogy of using a set of bathroom scales -- do you choose a set that always reads low, or always reads correctly?