1. Statistical analysis is good, but claiming 100% accuracy is unrealistic. It is an educated guess at best. Ring Signatures offers plausible deniability. There will always be a degree of uncertainty when trying to model a selection algorithm that picks decoys from the entire blockchain with a distribution pattern similar to normal user spending habits. Even if a deterministic transaction graph is made of all outputs breaking Ring Signatures in the process, RingCT and Stealth Addresses still de-links outputs from transaction amounts and receivers' wallet addresses.
2. No one uses Zcash.