A few problems here:
1. >flows to addresses which have been identified as illegal
These are only identified as illegal. First off, only a small number of adressess has been marked as illegal, since it doesn't make much sense to have everyone know the dirty deals you're making, most of the real illegal traffic is not there. Also a large number of DN markets uses 1-time wallets, so it's impossible to identify those.
2. Most of the traffic analysis is based on private to/from exchange and exchange to/from exchange data. As exchanges require a KYC, it can be assumed that most of the exchange routed transactions aren't illegal, or first go into a mixer if they are coming from an exchange, decreasing the odds that it can be classified as illegal.
3. Analysis was done in a crypto bubble market, where everybody and their moms were trading it on a daily basis to earn money.
In conclusion - the paper is bull. Yes, a large amount of usage is speculation, but illegal activity is way larger than assumed in the paper. But also, when you remove trading/gambling for speculative profits, what usages remain?
Mostly illegal ones, small amount of legal ones and a few believers.