It looks like the 90m cumulative times are off. That screws up the splits before and after. I had an assistant doing the splits so I'll look at the raw file and see if the real data is there.
The only thing suspicious about Gay's race is that the last split is the fastest. What else are you seeing?