I remember this paper. A major flaw that was pointed out to the authors (and never corrected) is that Bitcoin tumbling services—popular among illicit users—artificially increase the transaction volumes. So if tumbling services puts the BTC on average through 10 transactions, then this will multiply by 10-fold the volume. In other words there wouldn't be $76B transacted annually, but only $7.6B.

Tumbling also makes some of their other figures unbelievable. For example in table 2 they report "181.82%" of all users transactions attempt to use tumbling services... I have little confidence in this paper being accurate.