AM say 100% of the old was traceable and up to ~50% of the new transactions are traceable.
He does not say that. He never claimed 100% of anything (though some high numbers like 80-90% on early transactions; see Table 2 on page 6 of the paper), and the paper says this about current transactions:
The weakness studied in this section is pri- marily a concern for transactions made in the past, as transactions using the new RingCT transaction option are generally immune.
RingCT was activated in January and essentially 100% of new transactions are now RingCT (stats here:
http://moneroblocks.info/stats)
Even before RingCT the vulnerability was plummeting due to earlier improvement (mostly mandatory minimum mix factor). See gray line on Figure 5, page 6.
These issues have all been widely known, widely discussed, written up in at least two papers, and the motivation for two widely-publicized hard forks. This is all a rehash and FUD pushed by Zcash.
EDIT: Oh I see you are misinterpreting his quote:
The first one leads to “conclusive” linking like we can tell with 100 percent certainty that a particular transaction is linked to another.
This method only applies to older transactions.
He
does not say this applies to 100% of old transactions. It is 100% certain as far as tracing
when it applies, but that is only for a portion of those old transactions (%ages given in the table I mentioned above)
Nor does he say that 50% of new transactions traceable (in fact none are according to the methods of the paper). The paper states that because coins are more likely to to be respent in a short amount of time, the most recent output in a ring signature is
more likely to be the real one, but you still can't conclude that it is. Often it won't be. These are still untraceable. This issue is also long known, partially mitigated and continues be improved, and this too is acknowledged in the paper:
Each subsequent update to Monero’s mixin sampling procedure has improved the resistance of transactions to the Guess-First heuristic.