I think this article really deserves a bold clarification/correction. It uses “transaction volume” in a few places as a synonym for “total quantity of ETH sent”, which is not a usage of that term that I have ever seen in ethereum. Phrases like “the share of the Mixer in all Ethereum transactions has changed over time” and “an overall growth pattern of Ethereum transactions looks very differently when the Mixer share is excluded” contribute further to the misleading nature of the claim, as nobody ever uses “Ethereum transactions” to mean “quantity of ETH transacted” either; rather, such phrases refer to the number of transactions sent (https://etherscan.io/chart/tx). This use of terminology is already confusing people, see https://twitter.com/zooko/status/911427963089850368.
Furthermore, I think that “total quantity of ETH sent” is a completely useless metric, because it is easily spoofable; if I were to send 400k ETH to myself in every block, then that would generate ~2b ETH moved per day, outshining whatever this “mixer” is by a factor of 100, and I would probably be paying less than this thing in transaction fees to accomplish that. This is precisely the reason why the bitcoin community prefers talking about “coin days destroyed”.
This finding is substantial, but it’s a VERY different result from what it’s being misinterpreted as, which is something like “2/3 of all activity on ethereum is one single mixer”, which is NOT true.