Probably a case of simple geometry of the BIG picture as opposed to a microscopic/focused look at the system.
At any individual point, 0.002" isn't much. But when you do some math over larger lengths, you see some interesing magnification issues become quite apparent.
0.002 clearance is 2 parts in 1000. Or zero clearance at X=0.000 and 0.002 at 1".
OK, now go out 12 inches. O.000" at X= 0.000, at 1" it is 0.002 but at 12" it beomes 0.024" - simple proportion when you use the law of Tangents.
What you have to do is slowly start lapping the gibbs over the full length of the table. As you remove the high spots and local deviations, you'll reduce the clearance without encountering localized binding.
Long steady process that takes time, care, patience, fastidious attention to details and strong, durable arms.