Late to the game here. But, if you study any of the engineering forums you will learn that measuring fastener tension (stretch) via twisting forces (torque) is highly inaccurate. The ballpark accuracy, you will discover, is about +/- 30% due to many variables, including fastener condition (thread condition, roughness, friction, etc.,) user skill, and on and on. This 30% comes up over and over again in various engineering discussions.
So, the difference between +/-2% and +/-4% becomes somewhat a moot point. Kind of like the difference between using a tablespoon and teaspoon to measure the volume of water in Lake Michigan.
Don't believe me? Study it yourself, then come back here and prove me wrong.
Here's an easy to read primer:
http://www.surebolt.com/torque_errors.htm