I tried this game and I completely disagree with your goals.
http://tasvideos.org/userfiles/info/63668734223205509
----------------------------------------
| | feos | link |
----------------------------------------
| | frames | igt | frames | igt |
----------------------------------------
| 1 | 442 | 09.46 | 519 | 10.60 |
| 2 | 532 | 10.84 | 601 | 11.77 |
| 3 | 543 | 10.90 | 599 | 11.68 |
----------------------------------------
The table only compares the hurdles, because dash is completely trivial. It counts from timer start to timer stop (for the winner). My movie also reaches the 4th dash 10 frames sooner than yours, but I don't think it's relevant. What's relevant is that making sloppy hurdles as a goal is extremely arbitrary. It's like playing Sonic in real time without rerecords and then reaching the level end at a point where it gives you the lowest score, so the next level starts sooner.
It's not like there's nothing to optimize in this game. Reaching the finish line ASAP is really hard due to unpredictable nature of jumps in this game. So it's a real challenge and it involves serious TASing. Hitting every hurdle and wasting time just because the opponent is still running is not a speedrun goal. It's technically meaningless, and also not entertaining, because it's visibly sloppy.
The amount of time I was able to save even in your first level, which you said you've optimized, shows that even by the stated goals this run is suboptimal. And after having tested this game myself, I'm convinced that aiming for overall movie time with this game doesn't lead to a meaningful speedrun record either. Because it's borderline trivial.
I tried to figure out how the timer works, but it's kinda complicated. I still think in-game time as a primary goal makes perfect sense (even if you manipulate it a little bit?). But ideally, one would optimize for race duration in frames, which is what we consider actual gameplay in this game, and it's what we always compare when deciding on obsoletions.
This run will have to be rejected.