Observations on HF clicker torque wrench

I was asking about the HF clicker torque wrench recently ("Cheap click-stop torque wrench" - 9/18). I bought one and just tested it, using a lever & weights:

The lever is a 1" square tube, with a sawed-off 3/8" socket welded to one end. The coffee can has lead weights in it. I gradually move the weight out 'till the wrench clicks, measure between the center of the socket and the weight-holding wire, and multiply that distance by the weight.

With the wrench set to 20 ft-lbs, it clicked at a torque of 17.4 ft-lbs. 13% low. Set at 50 ft-lbs, it clicked at 43, 14% low. HF claims that it's accurate to 4%.

Also interesting is that it has memory. I.e., it acts differently depending upon what it has just done. E.g., when testing the 20 ft-lb setting, it clicked when I got the (8.54 lbs) weight out to 22". But if I then moved the weight back, it would click at shorter distances (lower torques), until 20-7/8" (16.6 ft-lbs). If I let it sit a bit & started over, it didn't click until 22" again. It loosens up when it clicks.

I was tempted to test whether this was a matter of the viscosity of its grease by seeing if temperature affected the click point, but I wasn't that curious.

I dunno if I should send it back. Maybe my testing is not as accurate as I think (although there's not that much room for error). I can pretty easily add 15% to the torque setting (e.g., set it for 23 when I want 20). Opinions?

Thanks, Bob

BTW - those of you who noticed that 8.54 lbs x 22" is not 17.4 ft-lbs, it's because there's also 1.79 ft-lbs of torque from the weight of the bar.