That's not a response I expected especially when it is concerns to your TAS but with that being written the input is definitely not optimal (I'm ignoring the naming case).
I need to make this note to everyone. The factor of the dipswitch is only related to the qualifying time. Not how the input is affected by it with that being written and noted.
Your time on the 100m sprint is 5"73.
My time is 5"68.
For me I had to use libretro's core in BizHawk in order to get a result for myself as I'm unable to TAS anything far with MAME-rr and I'm fine with waiting.
But I'm still able to show that there is a problem with your input and luckily enough I'm also provide
input that syncs (for today's core).
Note: I do not recommend TASing with any of the libretro cores within Hawk unless it's for experimental usage as I've demonstrated here.
Edit:
Your record breaking Long Jump of 10.78m was improvable.
Your time on the 400m Relay is 30"70.
My time is 29"76.
I want to also take the time to point out that the AI got off the starting blocks before you did in your TAS.
Here's
a file that syncs both the longest Long Jump and the Faster 400m Relay (and the 100m sprint).
This is the last event I'm touching as I don't want to deal with Skeet Shooting and this method of TASing on something very experimental.