CSV output files from Psychopy - muddled times

Hi everyone! I was wondering if anyone had experienced something similar or had any advice:
The CSV file from my Psychopy experiment provides times on and times off for the stimuli presented on-screen, e.g. for an region of interest (it’s an eye-tracking experiment). Looking at these raw times given, the ‘time off’ is consistently before the ‘times on’- so it’s saying the region of interest came on at 3.94085 ms, but switched off at 3.4727984 ms.
Another example- region of interest time on at 3.4725296, and times off at 3.2719858.

Does anyone know why these times are flipped like this, and whether I can simply take the values as swapped? Thanks for any help! Chloe