Does sen2cor needs whole image for correction

I’m using sen2cor as part of chain in which the final product is a resampeled subset of the original image with all the bad pixels (defective, saturated, cloud) are set to 0 on all bands.

Is sen2cor pixel-specific or it uses the whole image for correcting
In other words, does removing (zero out) pixles and subseting reduces the correction quality

For context, currently my chain is:
download -> sen2cor (if it’s a level-1C product) -> resample -> subset -> remove bad pixels
And I want it to be:
download -> remove bad pixels -> subset -> sen2cor -> resample

Thank you

sen2Cor needs the unprocessed L1C data, any changes in the input are likely to cause the process to fail.