Good morning,
Thanks for posting that table. This now that we know reads are getting lost at the merging step, we can look for settings that will let more reads be joined.
I understand the time pressure of group meetings, so let's dive in! 
Yep, something is going poorly at the merging step. Having 20 bp for overlap should be enough, so let's figure out what could be going wrong.
I'm not sure... the number of reads 36k or just 300 reads will not effect overlap, but the length of the amplicon and length of reads will effect that. So if your amplicon is 250 bp long and your reads are 150 long, you will expect 50 bp overlap. Like this:
250 bp amplicon |-------------------------|
150 bp read |--------------->
150 bp read <---------------|
50 bp overlap ^^^^^
However, if your forward or reverse read is low quality, then the reads will overlap, but dada2 will be unable to pair them.
250 bp amplicon |-------------------------|
150 bp read |--------------->
150 bp read <000------------| (30b are low quality)
50 bp overlap (with errors) ^^ (so only 20 can join)
One solution to this problem is to trim the end of that reverse read so that the part that's left is high quality and is able to join.
Maybe trimming will help! Try running qiime demux summarize and look at the Interactive Quality Plot tab. This will show you the quality of both forward and reverse reads and you can use this information to see where you need to trim.
While you can make OTUs or ASVs using these reads, having 36k reads would be much better! Let's see if we can find some settings that let you make use of your full read library.
Let me know what you discover in the quality plots! 
Colin