There are two things to keep in mind when watching videos recorded through the HoloLens:
1. The frame rate and image stability are much better in person. The HoloLens has to cut the rendering frame rate in half to be able to record video through its front-facing camera because the camera and the display share a bus. So what you see is significantly degraded.
2. The constrained field of view is much more noticeable in person. With the video, you don't get to see anything out of the bounds of the frame, so there is no reminder that we can't "see" everything. Also, we are used to watching videos, so we don't think about the fact that we can't turn our eyes and look in different directions. What you see is significantly more generous to the HoloLens than reality.
I think the combination of the two factors generally makes the video an "accurate" representation of the HoloLens, even if it's not terribly "precise".
I don't consider the current iteration of the HoloLens to be a usable device. Comparably, the HTC Vive is a much better experience, one that is capable of providing to users an experience that is worth the drawbacks and the extra effort necessary to deal with them. The HoloLens is more effort than it is worth.
But on the other hand, tethered, completely opaque systems that require external tracking hardware for head and limbs are clearly a dead-end technology, even if that dead end is 5+ years from now. The end goal with mixed-reality technology is clearly in the HoloLen's lineage, not Vive or Rift or even Meta 2.