@BBCRD Interesting read. Could it be that the computer you used for testing simly didn't have as optimized hardware/software?
You claim that the difference is from different measurement methods (I agree that HDMI capture is a better "ground truth" metric), but it doesn't seem you actually tested requestAnimationFrame().
If requestAnimationFrame() testing showed that the WebGL framework was slower than the DOM framework, then you couldn't outright dismiss the WebGL framework's claims.