You have sequenced a region of the 16s gene that spans from base number 341 to base 785 (from your primer names). If you take the difference of these sites you get 444bp which is around the length of the fragment you are getting so it all makes perfect sense.

The v3 kit from Illumina generates 600 bp of Paired End reads by generating a 300bp forward read and a 300bp reverse read. If your target region is larger than 300bp then these forward and reverse reads will not fully overlap and therefore your contig will be longer than 300bp.

This is a relevant read.

In terms of picking min/max lengths for the screen.seqs command there is no correct answer. Variation in the 16S gene means that its not unreasonable that sequences larger than your mean length are legitimate, but on average the longer sequences are usually trash. Pick a value maybe 10% greater/less than the mean as a starting point and compare to something like a 20% greater/less than the mean and see how that affects your results. I found once I was trimming too stringently and losing some archaeal sequences.