Hello @Shreya,
EDIT: I've rewritten this post. I hope it's easier to follow.
If you choose to use DADA2, trim where the quality drops to maximize total reads passing through the DADA2 pipeline!
This thread has some good advice.
I usually run DADA2 with a few different truncating settings to see what works best.
Because DADA2 builds an error profile for each sequencing run, we recommend processing each run separately.
Once you have DADA2 results for each run, you can choose how to combine them. This is pretty easy if all your reads are from the same region!
Because you have data from different regions, this may be tricky...
Do both runs fully cover the same part of V3?
Or it may be impossible! 
Working with multiple regions is a big challenge, so let us know if you have more questions!