From Null-Structural Applicability to Fixed-Reference Admissibility: An Integrated Framework for Conditional Physical Extension

Foundational proposals can move too quickly from a semantic boundary to an ontological commitment, from ontology to physical relevance, and from physical relevance to a proposed experiment. This article gives a single type-safe framework for auditing that chain in a null-structural case and for stating one conditional physical extension without presenting it as a consequence of the underlying semantics. The first layer separates domain applicability from truth: a structural value may be evaluated only under a live domain guard, so inapplicability is neither falsehood nor a zero value. The second layer uses scenario typing to distinguish a denoting Scenario-A NullSector, a non-denoting rejection branch, a first minimally structured sector, and a non-denoting boundary surrogate. Scenario-A actuality is modal and typed; it does not imply physical presence, location, interaction, or observation. A preservation result labelled SURVIVES is correspondingly restricted to framework-relative non-elimination. The third layer separates formal actuality, ontic commitment, referential warrant, and operational physical relevance. Shared foundational content cannot discriminate a strong referential reading from a live structural-realist comparator, while positive target-internal enrichment fails to preserve a target constitutively specified as null-structural. The resulting warrant condition is necessary only: any successful warrant must add independently justified, discriminating information without replacing the target. The fourth layer preserves earlier non-entailment results: bare ontology, a nonempty ontology-to-physics interface, and structural or dynamical extension status do not automatically determine physical relevance or observation. Against that firewall, the article introduces two independent declaration premises. H_dir supplies a well-typed opposite-resultant direction-selection rule, while the separate instance predicate D_elig records whether a declared uncertainty set actually passes its direction certificate; H_ctrl declares the model, policy class, authority predicate, and typed tolerances needed for a control test. In an explicit local Fermi-normal and small-body regime, Proposition A gives the exact angular diameter of a Euclidean ball uncertainty set, Proposition B classifies fixed-reference and adaptive protocol histories, and Corollary B1 identifies the evidential loss produced by a target-blind observation projection. Lemma D supplies a sufficient finite-size reduction bound with curvature, non-gravitational, and carried-frame first-order terms kept distinct. A retargeted success belongs to the adaptive class and cannot be counted as evidence for the frozen target. The framework is therefore operationally rejectable under a declared contract, but it supplies no V₀ detection, transition law, preferred frame, or experimental claim. This is a public preprint and has not been peer reviewed.