My comment about the false alarm is directed, naturally, at the title of the thread, and as far as the title is concerned, I don't think there is any need to equivocate.
But you have a point anyway, because it is possible that Jim's tests might have come out differently, in which case the fact that a relatively simple variation of the correct knot would have been catastrophic.
Hence, the video highlights the fact that it could be very dangerous to modify a known knot in a way that might seem intuitively advantageous. We have already seen this play out, tragically, with the substitution of a figure-8 knot for an overhand knot in the EDK. A second example is that at least some so-called Yosemite finishes, when added to a figure-8 loop, make it more prone to rolling.
The fact that a bowline does not remain after the transformed version is partially untied seems to me to be an artifact of terminology more than anything else. The bowline with Yosemite finish is a singe entity, regardless of its compound description. If you rearrange it and then undo part of the rearranged version, it shouldn't be much of a surprise if the starting configuration is no longer present.