The meeting scenario displays eight thousand four hundred milliseconds for one architecture and one hundred ninety for the other. Dividing the authored constants gives approximately forty four point two. At point zero four pixels per fixture millisecond the bars span three hundred thirty six and seven point six. This page does not call either provider or time a real task. Its steps and claimed capabilities are fixed scenario content. A button labeled Run Trace animates that content; it is not evidence that these products achieve those latencies or handle the stated applications. The actual trace animation uses a separate clock. Four steps in the native-labeled loop each wait sixty milliseconds, totaling two hundred forty programmed milliseconds. Four steps in the other loop each wait three hundred eighty, totaling fifteen hundred twenty. At point two pixels per programmed wait millisecond the bars span forty eight and three hundred four. The loops run sequentially, with browser scheduling overhead added. Their timing therefore does not demonstrate the displayed one hundred ninety and eight thousand four hundred millisecond benchmarks. It shows a designed visual replay, not measured parallel agent work. The architecture-fit calculator uses a hand-authored points rubric. At a latency requirement of five hundred milliseconds or less it awards four points to the native-labeled option. Moving the slider to five hundred fifty awards only two points. At eighty pixels per latency-rubric point the bars shrink from three hundred twenty to one hundred sixty. Other controls also add points, so this change alone need not change the final recommendation. These slider values do not modify the meeting trace constants or its exported forty four point two ratio. The calculator and scenario replay are separate mechanisms, and neither validates the product assertions on this page.