Me, too, on not using those light tokens. And, I agree about averaging out on the weight. My error rate runs about 2% max.
The difference in weight is negligeble one on one. But if they are the exact same, when weighing a lot of tokens than there in no error rate. :-)