We introduce GAVEL (Grounded Adjudication of Variations across Extracted TimeLines), an LLM‑based judging protocol that compares two extracted clinical timelines against their source case report. For each discrepancy GAVEL returns a discrepancy type, a verdict, and the relevant passage from the report, enabling fine‑grained validation of extraction outputs.\ \ In our evaluation we first measured the performance of the event matcher, then reviewed 2,738 findings produced by GPT‑5.6sol and DeepSeek V3.2. Using these results we ranked six LLM extractors and two human annotators, and we tested a GAVEL‑guided timeline merging workflow.\ \ True match rates were 60% just below and 48% just above the 0.10 cutoff. Manual review confirmed 89.4% and 88.6% of the findings respectively. Across 126 reports, merged timelines were preferred in 77.0% of pairwise comparisons (95% CI, 69.8%–84.1%) and reduced the average number of discrepancies per report from 7.63 to 0.85.\ \ The key benefit of GAVEL is that it treats neither timeline as ground truth; instead it directly aligns timelines with the case report, providing a more reliable assessment and iterative improvement loop for clinical timeline extraction.\ \ Review