Generative Structural Theory v1.17 — Bijective Truncation Behind the Fixed-Census Separation

GST v1.14 found that, on a fixed single-arc census record, the provenance-population condition Q_D selects a strictly smaller set of candidates than the recorded verdict condition E_rec. GST v1.15 characterised this difference conditionally through the fibres of a truncation map from full chain tuples to prefix chain tuples. GST v1.16 recomputed the relevant counts and established that both candidates in the difference set have injective truncation maps, leaving their surjectivity unreported. This version determines surjectivity by extracting and cross-checking values already computed and published in the v1.16 supplement. No new counting is performed. The extraction procedure, checks and target population were registered and locked before the extraction. For both candidates, arcs 4 and 24, the full chain set, the image of the truncation map and the prefix chain set each have 952 elements, with branching surplus b_a = 0 and empty-fibre count h_a = 0. The truncation maps are therefore bijections of finite sets: every prefix chain tuple has exactly one full chain tuple extending it. For each of the three extended letters, all prefix chains end at a single live end point with exactly one admissible extension. The per-letter counts are 7, 68 and 2, whose product is 952. Thus Q_new fails by equality of the full and prefix counts. Q_D is also false for these candidates, consistently with the v1.16 set comparison; this does not establish the record premise PR. The result is restricted to the two candidates on the fixed inputs. The structural reason for their unique extensions remains undetermined. The supplementary archive contains the extraction code, registered inputs, synthetic tests, execution outputs and provenance records.