Hi @ninaxhua,
To be honest, I haven't seen any published meta-analysis with DADA2 that combine different primers and sequencing technologies but that doesn't mean that they don't exist or it's not possible to generate them. Anyway, in my personal experience, we normally use close reference but in recent months we are moving to use deblur. In fact, the fragment-insertion tutorial is an example of deblur meta-analysis combining different regions.
Well, in theory this should work but in praxis I haven't seen great success using close-reference. I believe this is due to the primer biases. However, I don't know if anyone has actually done this test with either DADA2 or deblur, perhaps worth checking, if this is something you are interested in.
I'm not sure I follow this question. Anyway, the primer normally is in the forward and sometime in the reverse read; however, this depends on the sequencing protocol (not the sample preparation where you add the primers). My suggestion will be to ask your sequencing center to be sure. BTW a lot of times is pretty easy to see if your primer is in your sequences cause you will see them "clearly" once you inspect your sequences.
Basically, fragment insertion is used after you get your denoised sequences (produced by DADA2 or deblur) and this seems to reduce the effect of the different primers at least based on the tutorial linked in my first answer in this message.
Hope this helps.