Per some of the inputs; I am in agreement with D. Hoover,  nailheading has
nothing to do with wedge voiding (also called chisel point voids).  The
circumferential voids that were the point of discussion was the possibility
of the direct plate coatings such as the carbon or graphite which must be
microetched and  undercut and removed from the copper surface interfaces
before additional plating; the microetch may etch away a small edge at the
extreme of the nailhead and leave a ring of non-conductive surface which may
not plate.  Several users have said that they have not seen the problem when
using this process.  The wedge voids are a completely separate
cause/condition.

Several people have noted that different copper foils such as the Grade 3
(HTE) created more nailheading than the Grade 1, but that the Grade 3 was
better for multialyer boards.  Many of us have seen this condition and even
some brands of HTE are more prone than others.  The metallurgist will tell us
that the more ductile foils where the slip planes (111) in a large single
grain are lined up will give us more metal flow and nailheading.  The rolled
annealed copper used by the flex fabricators is metallugically perfect for
providing the greatest nailhead.  It is noted that IPC-RF-245 (1986)
specifies nailheading can be up to 2X.

John Andresakis in the April issue of Circuit Tree (Page 24) reports that the
fine grained foil that he has tested has a much lesser nailheading tendancy
than the HTE.  He also notes that HTE (Grade 3) nailheaded more than standard
electrodeposited (Grade 1).  Again the metallurgist will tell us that the
fine grained copper foil with unoriented slip planel is less ductile and will
flow less than that large grain material and decrease nailheading.

I think the best solution is drop all requirements for nailhead as a defect
and let the fabricators make their drilling/processing etc.
compatible to produce a good board that meets the performance requirements
without this limitation that is not function related.      

If the fabricator so desires, he can set up a "process indicator" that reacts
to certain limitations on nailheading that may be material/condition
dependent and proceed from there.