As far as I know there are no false positives or negatives in my data. If anyone's interested in the full data (27 GB of "which chip exists on which day"), let me know.
Do not you think the amount 0.001 or 0.002 could be so popular that connecting it with chipmixer transfer is a bit risky? Or course we may say it is a candidate because that amount was used, but I think we may have many many uncorrelated transactions.
How about source addresses, any duplicates?