https://www.theverge.com/2019/2/15/18226105/a-note-about-youtube-copyright-strikes-from-vox-media
That right there is factually wrong. Maybe they are counting on nobody actually doing the math. Let's see Bitwit's parody video, it's 9:12 long (552 seconds), whereas The Verge's video is 11:10 (670 seconds). Even if Kyle's video was showing the Verge's footage non-stop throughout the video that would amount to 82.4% not 90%. But not even this is true.
In the first 11 seconds of Kyle's video he uses footage from The Verge, then it returns at 00:16s (-5s then) with a still image that stays on screen until ~00:19s. Then there is the sponsor and some commentary with only Kyle on screen and the Verge's video footage only returns at 00:47 (-28s then) and it shows a variation of a still image from the Verge's video stopped at 00:11s until 00:53 on Kyle's video. Then, from 00:54 to 00:58, it's only Kyle on screen commenting (-4s). Then, from 00:58 until 1:02,
it's the same still from the Verge's video stopped at 00:11 (which basically has only "How to build a custom computer" written on the screen against a white background). Then, at 01:10, Kyle pauses the Verge's video at 00:19s and it stays like that until 01:16. Then at 01:24 he pauses the Verge's video again at 00:27 and it stays like that until 01:31. Then there's the infamous "tweezers moment" at 00:37 in The Verge's video and Kyle pauses it there at 01:41 in his video and talks about it until 01:56.
I could go on and on, but let's just jump to the end: Kyle stops showing anything of the Verge's video at 08:11, so a full minute and one second (-61s) before the end (at 08:28 he shows the Verge's YouTube page but you can only see a tiny bit at the top of the actual video (paused), what he is showing is that they had disabled both the likes/dislikes and the comments on the video, so it doesn't really count).
You see where this is going, right? Let's jump over the fact that Kyle paused the Verge's video numerous times and thus he couldn't possibly have used even 82.4% of the thing, which is the absolute maximum possible given the lengths of the two videos, let's just subtract all the times where the Verge's video isn't shown:
552s-5s-28s-4s-61s=454s = 67.8%.