Every input needs a date.

Heirok: preserve the discrepancy.

The 2014 index links AC33 Heirok to 34.pdf, which is also linked for Wangjing Tentha. The original 34.pdf header reads 34 - Wangjing Tentha. A separately retrieved 33.pdf header reads 33 - Heirok.

The original index reference and diagnostic recovery remain separate in the audit. No duplicate 34.pdf vote table is silently assigned to Heirok, and no numerical values are promoted by the header check alone.

Inspect the recovered Heirok header ↗

Earlier context, not a new target.

2014 Parliamentary observations could eventually provide pre-2017 context, and 2019 could provide pre-2022 context. They remain Parliamentary results and cannot replace Assembly election targets.

Remaining work: broader 2014 individual scanned-sheet checks, resolution of the 2019 Henglep contradiction, and historical source-availability evidence. No fitted model, 2027 likelihood or complete all-page review is claimed.

Henglep: a 293-vote contradiction

The segment footer prints candidate columns 3, 5 and 7 as 924, 17,473 and 196. The combined Parliamentary sheet prints 944, 17,743 and 199. The segment candidate cells sum to 25,678 while its printed valid-vote total is 25,971.

Three polling-station rows also fail candidate-sum accounting: 57/7 (176 versus 196), 57/21 (535 versus 538) and 57/47 (152 versus 422). Their deficits total 293 votes. Both source versions are preserved; no replacement is silently selected.

Inspect the conflicting combined sheet ↗

What the checks establish

All 32 Inner segment totals reconcile candidate by candidate with the combined ordinary total. The 27 released Outer tables plus the separately transcribed combined Henglep row reconcile with the Outer total; that cross-check does not resolve Henglep’s conflict.

The combined sheets explicitly label postal ballots. Ordinary plus postal equals the grand total for every candidate and ballot category. No postal votes are allocated to individual Assembly segments.

Moirang’s printed station ID “17/15” is preserved under the PDF’s AC27 header. Thoubal 31/24 column 8 has a selected visual zero-with-stray-dots correction, recorded with its page locator. Selected visual checks are not an independent human audit.

Record when the event happened.

A 2024 Parliamentary segment observation cannot enter a 2017 or 2022 prediction. Its election type remains Parliamentary context even when its geographic unit has an Assembly constituency number.

Record when evidence was captured.

Date captured by this research snapshot; not the first source publication date. A later capture does not prove historical availability. Earlier source publication/archive timestamps are not independently established here.

Keep observations separate from effective dates.

The directory columns record the 2026 capture. They do not certify membership on a different date. A listed vacancy has no party value; it is shown as not applicable when its observation is within the cut-off.

Keep missing inputs explicit.

Complete verified 2027 nominee sets, verified alliance assumptions and reviewed representative opinion inputs remain unavailable in this register. That is a coverage limit, not a claim that no announcements or studies exist.