False Knock?

Has anyone had issues with false knock on their EBM? I have a feeling my knock sensors are picking up something that's not really knock. There's a few reasons I'm thinking this. My Accessport will show negative corrections when lightly revving in neutral. The corrections will happen in any cylinder and are pretty repeatable. I know some random negative corrections are normal, but again this is repeatable. And I don't think the car is really knocking while revving to 2k rpm in neutral with no load on the engine.
I also will see inconsistent corrections at WOT. Some pulls will look flawless and others will have random negative corrections or inconsistent positive corrections.

I don't think it's bad gas - I added several gallons of 100 oct. to my last tank and this made no difference. Plugs were swapped out 2 weeks ago with pre-gapped Densos and torqued to 13 ft/lbs. My only thoughts are that something could be rattling in the engine bay or under the car that's getting picked up by the knock sensors. I've done a little googling on false knock and in most cases it seems like exhaust is the culprit. Another possibility on my car could be the UPR catch can. The threads in the bolt hole are jacked up pretty bad so the can itself is pretty lose and can be moved easily by hand. Before I go under the car and start checking the exhaust, has anyone had any experience with this or have any suggestions?

Mods are Mishimoto intake, Levels IC, Tune+ WGA, catless DP, Solo Performance catback, URP dual valve and clean side catch cans, and of course Accessport and 93 oct. tune.

Video below of the negative corrections while revving.

[ame]

I think it's random. Nothing to worry.

You should be concerned if you see 5-8 recorded knock count, every once in a while a 1 count is normal. I often under full datalog pulls will see it ping once but my timing never retards. Other times you can see a -1 on timing which is normal while blipping the throttle.

I would remove say cyl 4 ignition correction monitor and add like one knock count cyl 1 there instead and monitor it. 3 cyls is more than sufficient to gauge timing you do not need all four, you can infer what cyl 4 is doing by the other 3 or in a data log anyway...