When I was building models for customers on a full time basis , I tried the bubble level aand the plumb bob level setups
Checking the results with eyeball said these methds were not worth a poo poo.
So- I made a squared setup using four levels - these created a perectly square base referrence
I then measured up to the model from the squared beams
-resolution of 1/32" (.032) was easy.
But the setup was time consuming
basic problem being the "matched "parts of left and right wing panels or left and right horizontal panels ( the small wing on the back, according to some) were often mirror images and not matched left and right pieces
This condition is commonly caused by sanding techniques etc.,
Once the model is assembled -the result is like having one panel upside down.
The measuring devices see only TE and LE-
the observation technique described by Matthews and me -reveals thicknesses above and below the center line.

The eye can see extremely tiny variations - micrometer sized stuff - even old eyes with degenerative problems.
Place two long scales (one meter etc., ) on a table - supported on edges and spaced apart say a meter and such that you can sight along the tops to see if they are perfectly parallel.
-using a tiny bubble level taped on top one scale - adjust scales till visual referrence shows perfect parallel settings
Does th bubble show the critical changes of say a few thou ?
most bubbles do not
I have good quality bubble setups and tools from Starrett - brown n Sharp etc..
still- I find the visual referrence to be the most accurate.