Suppose you construct transaction with two inputs and one outputs. One of those inputs is tainted and very small in value (say a few sat) but the other one is a big and "virgin". Regarding output, should it be considered as tainted?
Should it be? Obviously not.
Will it be? Maybe.
We have no idea what metrics or benchmarks different exchanges, chain analysis companies, governments, etc., are using to deem coins "tainted" or not. Some will decide that that output is completed tainted. Some will decide it is clean. Some might decide it has some percentage taint to it, which if you deposit it to an exchange account, may rack up some red marks next to your account.
Centralized exchanges set these arbitrary rules, don't tell anyone what they are, expect users to blindly play by them, and then punish them when they trip up, usually through no fault or dishonesty of their own.
Well, I asked that because recently I read [
url=https://arxiv.org/pdf/1907.01538.pdf]one paper[/url] centered on analysis of tainted bitcoins authors of which have concluded that using small amounts of "bad" satoshies criminal may rotten the innocents addresses. Their thought sounds like this - "a criminal who knows how the approach works might just send tiny fractions of Bitcoins to random addresses to taint them" . So, criminals knowing that all their
BTC are tainted and having them in enormous quantity can intentionally contaminate as many addresses as they want in order to conceal their own addresses among all others.