Kyle said it- you're comparing apples and oranges. You've proven by your examples that you're not on the same sheet of music.
Also... Youtube is a make-believe land. Many impossible things happen on the Tubes... https://www.youtube.com/watch?v=NePCqRmQ-Ok
We do have some research on it, though admittedly, it's shoddy as hell (underpowered, mostly). It's TLR for me to summarize them all here, but a quick review found 9 studies from which A) tested the SVJ pre and post training and B) included the effect size or percent where I could get to it (ie: not behind a paywall). The highest increases were in previously untrained populations at 8.5 and 8% over a 4-month and 6 week intervention (respectively). Experienced athletes experienced increases of ~2-3% over 3-4 month interventions, and kids' SVJ didn't respond to the training at all.
If the question is, "Could a novice increase their performance on the SVJ test more than 20%?" Probably- I haven't met any, but maybe Will could elaborate? We know that weight loss, experience with the test, motivation of the athlete, and certain tricks do allow novices to 'game the test.' Combine that with training and I'd actually expect outliers, especially now that we've now got a myriad of 'elite weekend warriors' pulling every trick in the book to improve their scores on random circus stunts every other week (though, admittedly, they seem to prefer the box jump).
If the question is: "Is general neuromuscular efficiency trainable to a great degree?", my initial impression is 'no.' If NME was progressively improvable like the squat or the clean, I think you'd be seeing college/professional athletes with dramatically improved pre/end-of-career SVJs (at the same BW), but you don't. A collegiate/NFL strength coach won't see these Youtube-jump-gainz because his athletes are already motivated, already experienced with the test, and likely tried every trick they could find to bump up their initial SVJ in their efforts to make the cut.