Up to the mods. It was pretty tight. Framecounted using somewes.com, both before and after I clipped it from my recording.
Mod edit: Since this run is bufferable (that should be a word) then the best theoretical time should be 12.467s. Using somewes' frame counter, I get the 12.450s time for this run as submitted, but when using the actual frame counter on the emulator, it yields 2161 - 1413 for 783 total frames, which divided by 60 is 12.467s. This leads me to believe that a frame was skipped/dropped in the actual recording.
I'm accepting this at 12.467 for now, but I can do a deeper dive if you feel this isn't a fair solution. Let me know via DM. ~Char