unless you have microscopic eyes and the bubble is exactly the same width as the distance between the lines, there is always error. If the bubble does not fill the line gap exactly, you will always be guessing how far out of level the level is setting. is the bubble 1/16" from the line; 1/32"; 1/64"; etc.?
Obviously one would be best served by utilizing as level a surface as is available for my tests but the comparison tests will prove the level against itself and unless you have a certified level to check it against, that is as close to checking the accuracy as you can get.
If the bubble locations do not match, there is some part of the level that is not true with the others. Now, if you want to find out where it is, be my guest but at that point, I toss the level. It will never be dependable unless you were to have it machined and calibrated. Not worth the cost unless it is a very large level.
That's my point. I think:
based on the bubble being centered,
if it fails the drill bit test it is bad,
if it passes it is good.
The bits and shims provide you a reference surface.
You can further gauge how far the bubble moves vs. how level the level is.
As a sanity check I'll have to run some tests on this with my levels; 4", 9" and 24" plus the line level. The repeatability of the readings is one limit on the accuracy.
I guess if you have a perfectly uniform-density piece of wood or other uniform-density material that is floating, this
is a perfectly level surface but you should still do the end-for-end flips.