Sure,

I think a good portion of the variability in the error is due to the inability of the meter to actually vary by .01 GPM, instead it has defined steps that it reads out that jump a little more than .1 GPM.

A couple of my tests it would jump back and forth between the two, so I averaged those. In others it would hold constant, but because the steps are only .1GPM apart, anything that is within .05 or so GPM would be acceptable for a specific step.

There may be something to the barbs used too. I used 5/8" barbs which are good flow, but there is still a fairly abrupt step at the end of the threading, perhaps this added turbulence is increase the velocities along the wall perimeter.

Actually the more I think about it, perhaps the cold temperature of the water I used has a bigger role in the error of my results. The water used was cold tap water at probably 10C which has a dynamic viscosity around 1.31 (cp), where as a typical aquarium probably runs at a much warmer 26C with a viscosity of .862 (cp).

I should run the test again at a more typical computer water temperature 24C and a 5% ethelene glycol mixture to get a more accurate error representation to what we would see in a system. That's just alot of work...