The way you check a level for accuracy is to place it on a fairly level
surface ( level not being critical) and check where the bubble sets. Turn
the level around (180 degrees) and check it again. IF the bubble is in a
different place the level is not accurate.
This happens all the time.
max

I ran into a curious situation today. I have some trivial but significant
alignment problem with the left wing of my table saw. I was attempting to
gauge how far out of whack it is by comparing two levels.

For the first level, one of those orange deals with three vials in it. I
used this to level the saw initially. It reports the saw as being level
all the way around. If I move it out to the waffle wing in question, and
position it appropriately, it also reports the wing as being level all the
way around. Whatever difference there might be is too small to gauge with
this method.

So I thought I'd look at two levels simultaneously. I grabbed the head off
an old combination square, that I use for sundry low quality purposes. Put
it on the table, and it showed a huge amount of tilt. Put it on the wing,
and it showed the same huge amount of tilt.

Then I got the head off my good combo square, and it showed the same as the
first one. Grossly out of whack level wise.

I just don't get it. Why would one show perfect and the other show that the
left side of the saw is almost 1/4" higher than the right? That's a huge
difference. My first thought is length. The orange 3-in-1 level is about
3/4" longer than the head of a combo square. My second thought is the size
of the bubble. The bubbles in the combo squares are a bit larger, and the
vials are a bit larger too.

Just for kicks, I also tried with a bullseye level. It shows level all the
way around too.

Weird. No, there's nothing stuck to either of the square heads, no hardware
protruding; no reason I can discern why both of them show such a huge
difference from the other two levels.