Why " start=10368, end=25319"? Half of your reads finish before that, so they get dumped. An then those starting after 10368 too… You should look at your alignment, and decide from which base to which base is the right alignment.
Pick the first 20 lines of your alignment (you do not need to check the whole dataset), check the alignment in MEGA or similar, and compare it to the region you are trying to get. I have to say I am surprised to see different starting points… Usually those are very consistent once oligos are trimmed. You might have leftover bases from your oligos?
I think your start position is likely correct for your data. You might try using 25318 for the end position. It’s probable that your data aren’t great or that you have a lot of non-specific amplicons.
Also, are you removing the barcodes and primers from your sequences at the make.contigs step? We generally see longer reads like these for the V4 region when the primers and barcodes are still on the sequences.