And according to the Empire web site, some of their "carpenter's" levels are accurate to a slope of .0042" per foot. The carpenter's level is claimed to be at least as accurate as the machinist's level.
BCA Specs for slate:" The playing surface must be capable, either by its own strength or a combination of its strength and that of the table baseframe, of maintaining an overall flatness within a tolerance of .020" lengthwise and .010" across the width...If more than one slab is employed, the slab joints must be in the same plane within .005" after leveling and shimming."
If one takes the spec as stated, and not "a tolerance of +/-.020"...), then for an 8' (88"-90" playing length) table, .0027" per foot would be OK even if all tilts were in the same direction (the worst case).
So according to the BCA (and WPA) spec, one full line off on the Starrett (.0050"), repeated all the way along or across the table, means you are almost twice as bad as the spec. Things are usually more confusing than that, and things tend to average out somewhat.
I suspect an excellent table would measure half-a-line or less on a Starrett placed anywhere on the table in any direction. That would mean that even in the worst case, it meets the BCA/WPA spec, and almost certainly is much better.