120 — Controlled example confirms the small-record length ambiguity
September 5, 2026. Synthetic experiment, not new evidence of the writer's algorithm.

Result
Two different authenticated-encryption constructions reproduce every observed small-record binary length and corresponding Base64 length: AES-128-GCM and ChaCha20-Poly1305. Both used an investigator-chosen12-byte nonce stored before ciphertext with a16-byte authentication tag. This makes the ambiguity concrete: length compatibility cannot identify AES, let alone an AES key size or producing library.

Experiment
For each of the12 distinct knowledge/plan binary sizes, generated a synthetic plaintext28bytes shorter, encrypted it with each construction, prepended its nonce, and checked the resulting binary and Base64 lengths. All24 cases matched the target sizes and decrypted correctly with the investigator's own newly generated keys.
Separately generated standard Fernet controls at both endpoints of each of the nine conditional plaintext-length intervals in118. All18 produced the expected binary token lengths and round-tripped correctly. This checks the boundary arithmetic; it does not authenticate the real tokens.

Evidence boundary
No captured ciphertext was passed to an encryption or decryption operation. Captured metadata supplied target lengths only. Synthetic keys and ciphertexts were not saved or published. The constructed wrapper is a demonstration, not a discovered nonce/tag layout. This does not establish that either algorithm generated xz, nor that xz is encrypted rather than another opaque representation.

What would resolve the uncertainty
Writer code could show serialization, compression, cipher mode, nonce placement, tag handling and any additional authenticated data. An authorized key, together with the correct construction and parameters, could allow actual authentication and decryption. A public format specification or a provenance-linked producer example could also narrow the alternatives. Another entropy estimate or another length-compatible algorithm cannot choose between them.

Primary API documentation: https://cryptography.io/en/latest/hazmat/primitives/aead/
Fernet specification: https://github.com/fernet/spec/blob/master/Spec.md
Artifacts:120-format-ambiguity-control.py and private120/analysis.json. No source-site requests, code from the supplied archive, or search-bot wave.