It's probably best to process each Illumina run separately, so that each run can be trimmed with the best setting and denoised based on it's own error profile. Once you get to ASV tables, you could try merging them. (This also lets you detect and control for batch effects during sequencing!)
Note: This will only work if all reads cover the same region:
paired |--------------<-->--------------|
single |---------------------------->
Area missing in single end! ^^^
If you want to merge ASVs they have to match exactly, so you may have to trim them so they match.
If not, you could denoise both data sets, then compare your top ASVs with a MSA like mafft to see how well they line up, then reprocess with new trimming settings so the output matches. This method is more reliable, but it can take several tries to get the trimming just perfect so everything matches.