Coinjoin didn't use same size inputs, masking the ability to figure out which output belongs to which input.  Then the "change" for the sender is sent back to a one time use address, that can not be associated with the wallet it came from unless you had access to that wallet.

Still, there are holes, especially if people reuse addresses that they've made public and don't use DarkSend every time.  The biggest thing that will plug that hole in the future is mandatory DarkSend.  After that, Evan has something else going on that I know nothing about 

So yes, coinjoin as Maxwell described is possible to follow the money, but if DarkSend were used, it would be pretty much impossible to "prove" but possible to kind of guess at what is happening with enough analysis.  Evan says he has come up with a way to make it as anonymous as ring signatures, but without the bloat.

Speaking of bloat, Monero has solved the bloat problem?  Or reduced it?  I suspect reducing a huge bloat problem won't be enough, being that even bitcoin's blockchain has become ungodly huge.  Ours will too, but at least Masternodes can be given the assignment to hold it so that regular users don't have to, yet it will remain decentralized. 

Anyway, I still like the "mechanical" way DarkSend works over any kind of encryption system, because it obviously won't be long before we have more powerful computer,  especially quantum computers, and they will be able to decrypt the code, revealing all the transactions on the blockchain.  The DarkSend Blockchain will not be any clearer under the scrutiny of a Quantum Computer.