Is my torque wrench telling me lies?

Well, yes and no: A tale of two torque wrenches.

In addition to the approximately 12,500 rivets in the airplane, there are many hundreds of MS, NAS and AN fasteners, all of which are supposed to be torqued to spec using a torque wrench.  Over a lifetime of turning wrenches, I have managed to acquire five different torque wrenches, ranging in

size from a small beam-type, 1/8th inch drive wrench I use on bicycles to half-inch drive jobs I use on cars and trucks.  The two I have used exclusively for the airplane build are a Craftsman snap-type 3/8th-inch drive, where the desired torque value is set on the handle and a distinct snap is felt when the torque is reached, and a 3/8th-inch-drive Snap-On dial-type wrench (shown in the picture connected so as to compare torque readings).  The literature always advises that the wrenches must be calibrated in order to ensure accuracy.  I had always (erroneously) assumed that the accuracy would be "close enough" considering that all my tools are name brand and the torque specs for each size fastener are fairly broad.  Since the airplane is going to be hauling me and my braver (or more foolish) friends around in the sky, I figured the time had come to check this.

Considering how much Snap-On tools cost, I assumed (correctly) that it would be the more accurate.  I bought an electronic fish scale on Amazon and also bought a "certified" weight (accurate to +- 0.5%) to check it with.  Amazingly, the rather cheap fish scale was accurate to within 0.9%.  I used the fish scale to apply a torque to the Snap-On wrench (shown in the pic) and it seemed to be dead-nuts accurate, at least to within the error band of this rig.

The shocker came when I checked one wrench against the other: When the Snap-On read 100 inch-lbf, the Craftsman indicated 84 inch-lbf.  This is unfortunate since many of the torque values being measured are essentially impossible to read with the Snap-On.  Henceforth, I will apply a correction when using the Craftsman.

A dial-type wrench is required since the prevailing torque (the torque required to turn the fastener in the self-locking mechanism) can't be measured any other way.  Measured torque minus the prevailing torque equals the actual torque applied to the fastener.  Maybe I'm worrying needlessly considering the error which results from the varying amounts of lubrication on the threads due to handling the bolts with hands which are no doubt far from clean.  Clamping force is all that really matters, and can only be determined through measuring bolt stretch.