Based on the structured records:

AMP appears in exactly 4 experiment records — Batch 21 (reference batch), Batch 25 B ("2 gm AMP" charged in Step 3), Batch 30, and Batch 66 — all per the 9076 Expt Table.

No controlled AMP on/off comparison exists in the records, so no isolated causal effect can be established:

• Batch 25 B (AMP charged, 2 gm, Step 3, fed over 2 h at 60 oC): outcome is unknown — no application test report is linked to this batch in this drop. Its remark is only "After 2 hr feed slight turbidity /yellowish flowable liq."
• Batch 30 (linked to LBN 23): outcome unknown, no application test report.
• Batch 66: outcome not ok — "Results: Not OK (Gloss and Re-solubility inferior)" [S 9076 #66 not ok.docx, page 1]. But the report itself does not state that this sample contained AMP, so the failure can't be attributed to AMP either way.
• Batch 21 (reference): outcome unknown.

One document, LBN 49, mentions AMP alongside a resolubility/gloss observation ("matching resolubility in black, blue & violet 27 while slight reduced in gloss found in all pigments"), but LBN 49 is an unlinked STUB in the records and there is no control run to compare against.

Gap: No pair of otherwise-identical batches with and without AMP exists in this data drop, and only failed application test reports are present — so the records neither show AMP ever changing a result nor rule it out.