Yes, from my experience they are from eukaryotic DNA which for some reason has homology with the 16S primers.
There is a quality control step, which requires a Vsearch filter to remove anything that doesn't hit with the 16S 97-99% homology with 16S which has dramatically cleaned up my samples.
edit: The clustering methods in QIIME1 has a "sort of" built in filtering step. Remember the usearch in QIIME1 requires 97% match either open or closed clustering for it to "hit". DADA2 does not do this, so anything that doesn't fit still remains in the sample, but becomes unclassified.
https://forum.qiime2.org/t/too-many-unassigned-or-only-at-kingdom-level-features/293
here are my examples:
Prior to a quality control step:
After a quality control step:
Ben