I can pretty much guarantee you that it will have zero effect in confusing more advanced tools. I've tested way more complex and advanced things to try trick up analysis, and it's not easy. Sometimes even I'll momentarily fool it, but later it'll "back propagate" (correct term??) information from how the outputs are spent (and associated clustering), to get a better understanding of the transaction. Like I've seen them reliably determine which outputs are change, in settings that should be impossible.

Taking bustabit as an example, it does smart partial batching so it frequently sends transactions with: (1 payment, 1 change) and (2 payments, 0 change).  Naively they are indistinguishable, but in reality analysis software has proven to have almost no problems distinguishing once it's been able to collect enough information after they're spent.

---

If you want to trick analysis software, pretty much a prerequisite is reasonably uniform wallet behavior (now is a joke...) and good practices (e.g. avoiding address reuse as much as possible). This will create an environment where there's a lot less "redundancy" (??) in the analysis, such that it has to lean on increasingly fragile assumptions.  And then (and only then really) you can be cute and do something like a bustapay/p2ep  or import/export a reused address output from/to a friend or something.

Now they'll probably realize you broke their models, but it'll be too hard to figure out (short of having law enforcement contact you for help declustering ).


But yeah, if you just got two very strongly clustered wallets with different behavior and created a single coinjoin between them (even if it was undetectably a coinjoin...) it's not really going to get you anywhere against advanced analysis (although it'll confuse something like walletexplorer, which maybe is something you want to do).