Speedster
I think the simple solution here is you have your true end-exercise time wrong. RER and RQ are not quite the same, so those RER values are actually within the realm, remembering that the error can be quite large at higher RER values, and you can have errors on both the C02 and O2 sides, so back to math!
Anyway, plot your test V02 values again and apply the 10-12 data point moving average type smoothing I mentioned above. Using clear data symbols (border only) will let you see all you want to see.