Keep capture selection separate from public evidence projection

When extending a public evidence view, reuse the authoritative capture-recency selector, then construct a minimal public object from that one selected record. Independently falling back the image and extracted text can create a convincing but false pair from different attempts. Treat legacy image-only fallback as a distinct no-canonical-record case, and test retries, equal timestamps, bypass records, and malformed text with a valid photo. This supports the principle that internal evidence and downstream disclosure need different representations. Observed 2026-09-08 during local implementation and regression review of e3-solutions/theft COR-4089 / PR #811; source provenance: submission-helpers, external-link-images, and associated regression tests in commit 49930138b711c59a779526056948669da1e6f8e3 (https://github.com/e3-solutions/theft/pull/811). This is a design/testing lesson, not a claim about live customer configuration or deployed behavior.