Moving the discussion to the theoretical aspects, 5.12.6 considers AISI Steel Plate Engineering Data, Volume 2, Part 5 "Anchor Bolt Chairs" (they says 5, I think it's VII in my copy) as an acceptable procedure to design attachment of the anchor bolts to the shell. That procedure has in back the old work of Bjilaard evaluating the maximum local stress S.
In addition, 5.12.8 mentions evaluation of localized stresses in the shell as given in ASME Section VIII Division 2, Annex 4, using the allowable stresses given in this section for Sm.
So both clauses consider the sensitive of subject of "localized stress" in shell. By the other hand, "the hole" in shell gives its own stress troubles and it is worth to say that WRC 107/297 and FEA results seem to consider stress troubles at reinforcement edge.
IMO, when you are following a constructive rule as 5.8.1.2 and put the anchor chair outside the repad, you decouple possible stress issues because you may consider that under API provisions, the local stress in repad is not significant affected by the attachment. Placing the anchor chair on the repad means to accept to increase the local stress and you would be asked to document the actual level of local stress (which probably would be FEA work that must be accepted by your Client).