The mill base and lower column stub may well be integrally cast together, but quill alignment relative to table is influenced by the bolted faces. As an example just eyeballing the dimensions, lets say he had 0.002" of gap deviation under a bolt (one of the upper blue circles). That's about half the thickness of a sheet of paper. If the column is 6" wide, that's a rise of 0.001" per 3". So if he used say a 12" radius arm to check tram in X-axis, that would be 0.004" per side or 0.008" total dial deviation. If the assembly was perfect to begin with, but the column stub got bent by damage, or welding heat or whatever, all it takes is a tiny amount of deviation on the short distance flange surface to exaggerate into tram runout over a larger distance. Even a bit of dried paint or casting burr would do it. One can visualize that if 1 bolt was artificially shimmed & the other 3 tightened, you would have a different tram deviation result in X vs Y because of how the spindle is orientated relative to the bolts.
So the issue may be confined to the flange face as Rudy is suggesting which could be remediated with shims. But this assumes the table is perfectly flat & not giving a false reading. If the column & spindle was perfectly square but the table is bowed and/or twisted, the dial will pick this up in the tram test.
