Fix benchmark p95 calculation

scripts/benchmark_runtime.py

@@ -102,9 +102,10 @@ def _elapsed_ms(started_ns: int) -> float:
 
 
 def _summary(samples: list[float]) -> dict[str, float]:
+    p95 = statistics.quantiles(samples, n=20, method="inclusive")[18]
     return {
         "median": statistics.median(samples),
-        "p95": max(samples),
+        "p95": p95,
         "maximum": max(samples),
     }
 

tests/test_benchmark_runtime.py

@@ -0,0 +1,21 @@
+from __future__ import annotations
+
+import importlib.util
+import unittest
+from pathlib import Path
+
+_SCRIPT_PATH = Path(__file__).parents[1] / "scripts" / "benchmark_runtime.py"
+_SPEC = importlib.util.spec_from_file_location("benchmark_runtime", _SCRIPT_PATH)
+if _SPEC is None or _SPEC.loader is None:  # pragma: no cover - test setup failure
+    raise ImportError(f"Unable to load {_SCRIPT_PATH}")
+benchmark_runtime = importlib.util.module_from_spec(_SPEC)
+_SPEC.loader.exec_module(benchmark_runtime)
+
+
+class BenchmarkSummaryTests(unittest.TestCase):
+    def test_summary_reports_interpolated_p95_separately_from_maximum(self) -> None:
+        summary = benchmark_runtime._summary([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0])
+
+        self.assertEqual(summary["median"], 4.0)
+        self.assertAlmostEqual(summary["p95"], 6.7)
+        self.assertEqual(summary["maximum"], 7.0)