Let me give a little more detail, I'm not sure I made myself clear. Below is the table the system uses to determine if a pallet passes the weight check.

Loading Weight Acceptable Variance

0 – 17 lbs 5%

17.1 – 32.9 3%

33.0 – 81.0 2%

81.01 – 199.0 2%

199.01 – 499.0 1%

499.01 – 1,201 ½ %

1,201.01 – 10,000 1%

If I had to guess, I think these are based on the specs of the scale.

I would further guess the scale is quoted as good to something like +/- 0.5% +/- 0.5 lb.

At 17 lb, this would be +/- 17*0.005 +/- 0.5 = +/- 0.585

This happens to be just under 3.5%. So above 17 lb, the error is no more than (about) 3%.

At 33 lb, the error is has dropped to about 2.1%, so above 33 lb, 2% is a reasonalble estimate.

At 81 lb, the error has dropped to about 1.1%, so (if hy hypothesis is correct!) the error could have been dropped to 1%

At 200 lb, the error has dropped to about 0.75% (if hy hypothesis is correct!) the error could have been dropped to 0.75%

At 500 lb, the error is down to about 0.6%, so 0.5% as listed in your table) is back closer in line with my hypothesis.

At 1200 lb, my guess is that the specs for the scale say the error increases to 1% +/- 0.5 lb, so the % error was bumped back up!

*This is all conjecture, of course! Check the manual for this scale (if you still have it) and see if they give specs something like I conjectured. The numbers also don't match exactly with my hypothesis, so I may be totally off-base, of perhaps the person who made the table rounded strangely, or ...*
Tim F