You might try this method: use a set test time, i.e., 4 seconds: test in the area of each collar (1st wire location). If the readings agree, record the reading. If for example,one is 18 and one is 20, I average the two - record a 19 in this case. If the reading is wildly different from end to end, say for example 16/20, throw the piece out. However, here is another place to be careful. Sometimes the impression ball starts on a peak of a rough gouge area and then falls off during the test. A repeat test nearby the first test may indicate it is really a 16.
One can see at this point how important the gouged surface is. My machine generates a very smooth surface, yet occasionally I run into this with my cane - but, most gouged cane that I see on the market is not smooth and should be sanded first. If you lightly run your fingernail cross grain you will see what I mean. And, a very smooth cane surface will increase any read maker's success ratio.
As I said before, we should not expect results similar to hardness testing steel. Readings within within + or - a point are the best we can expect, in my opinion. I think you are on the right track - and I hope my comments will be of some benefit to you.