oh right of course! reading/writing too fast. 
same idea, though: the ITS primers are couched in highly conserved sections of 18S and 5.8S rRNA genes that flank the ITS.
Those look like very well-chosen settings, given the high quality of your data.
There is no reason to do that with your workflow. That link you provided was specific to that user's workflow — they are exporting their data and running through an external program before re-importing, so the filtering was done to remove dropped features if I understand correctly. (and an ITSx-like method ITSxpress is also available in a 3rd-party plugin now so there is no reason to follow that workflow any more).
could you please share the actual qzv files? this will be easier to peruse. Please also send a QZV of the dada2 stats. I suspect you probably have a read merging issue; the second link you sent above is probably the most relevant for troubleshooting, though your data quality looks very high.
One problem is that (if I recall correctly) ITS1F (fungi-specific, I'm assuming F does not just stand for "forward" and this is the standard ITS1 primer) sits pretty far back in the 18S, so your pairs might not be quite long enough to bridge the full ITS region. Do you want to figure out the expected amplicon length?