Electrolysis record merges two H2-supply arms into one setup; Caldicellulosiruptor note names the wrong species

[realmarcin]

Found reviewing PR #515. Both are in prose, both verbatim-snippet-clean, and both are the failure mode #513 already named: the snippet quotes correctly and the sentence beside it summarizes wrongly.

1. Acetobacterium_Clostridium_CO2_Electrolysis_Coculture describes one setup where the paper ran two. PMID:36619880 compares two hydrogen-supply arms in the same stirred-tank system:

Lactate concentrations as high as 0.5 g L-1 were achieved with the AiO-electrode, whereas 8.1 g L-1 lactate were produced with direct H2 sparging in a stirred-tank bioreactor.

The record presents a single cultivation_setup — a stirred-tank bioreactor with in-situ electrolysis — which reads as the configuration. It is one arm, and the lower-performing one, by a factor of 16. The direct-sparging arm is missing entirely.

This is the more serious of the two: cultivation_setup is multivalued, so the schema can already represent both arms. Collapsing them loses the comparison that is the point of the paper, and anyone reading the record would attribute the coculture's behaviour to a setup that produced 0.5 g/L when 8.1 g/L came from the other one.

It also makes a claim I could have sourced directly. I wrote that the paper "attributes the ceiling to the electrode" as an inference from the future-tense sentence about enlarging cathode area. The abstract states it outright — "Hydrogen limitation was identified in the AiO process" — which is better evidence and belongs on the AiO entry.

2. Caldicellulosiruptor_TwoSpecies_Hydrogen_Coculture names the wrong species. The note says C. saccharolyticus prevails at higher dilution rates. The source:

with C. kristjanssonii outgrowing C. saccharolyticus during the batch start-up phase and prevailing at higher dilution rates.

The subject of "prevailing" is kristjanssonii. I read the second-named organism as the subject — the same slip as the molasses direction in #513, one round earlier.

Both fixed in #515. Filed because the pattern is now three-for-three: every prose error this sweep has produced is a relation (which species, which variable, which arm) stated backwards or flattened, in a field no validator reads, next to a snippet that validates clean. Worth considering whether notes/instrument_detail need a review step that re-reads the source rather than the sentence.