Clock vs. component.rt

URL of experiment: /

Description of the problem: I am interested in the accurateness of a clock versus automatic logging. In my test, I am measuring reaction time needed to press the button upon hearing a specific word in an the audio file. So I have two options:
(1) tick “save onset/offset times” of the keyboard component and use automatically generated ‘keyboard.rt’, or
(2) start a clock in ‘begin routine’ tab and use getTime() in ‘every frame’ tab which is going to be triggered when the key is pressed.
I’ve tried both and it seems that logs from option 1 are 0.015 - 0.025 seconds (15 - 25 milliseconds) longer than option 2 which is RT generated automatically by the program. I want to know which one should be more accurate?
Also, if option 2 is more accurate (which I suspect), is there any way to access these logs in the program since I don’t want raw rts in the output file. I’d like to do some math calculations beforehand and only get the final value in the output file instead having to do that manually for each file. So I would do the calculations in the code component, but I am not sure how to access these data.
P.S. I know I can use excel macros to manipulate large number of excel files simultaneously, but I’d like to try within Pavlovia/PsychoPy first.