Hello Tina,
I'm still here. 
I think you are off to a good start. Having both a kit control and a known positive control is a great start, as it let's you set a baseline for what's going inside the context of your runs. The problem is still pretty hard...
I'm been thinking about what Nick said, all the way back in 2017:
Like, you mention,
Sure, you could get the ASVs from your negative controls, then remove them from all other samples using this plugin. But there's a catch!
The Illumina platform suffers from index-hopping between samples, and the most abundant amplicons are the most likely to 'hop' into other samples because of a mismatched barcode. This means that a perfectly empty negative control would still get some reads assigned to it by the Illumina sequencer, and essentially be a tiny subset of all your real amplicons. Removing these real amplicons is not a good idea.
The issue is that technical / artificial variation and biological / real variation all look the same to the sequencer. Sometimes there's recognizable patterns of technical variation that you can identified as noise, then removed. However, the real biology can be noisy too.
So, what do you think is the source causing your contamination? How can you tell it apart from real biological variation?