It may be worth investigating if indeed magnetically-coupled capacitor currents are introducing errors in the PT cabling. {As implied, non-ferrous shielding is primarily effective for electrostatic influence.}
If for instance, the PTs are used for capacitor-unbalance protection, some initial or quiescent voltage may be present from accepted capacitor-manufacturing tolerances, and somewhat influence what would be considered an otherwise near-zero secondary quantity, like on the order of a 0.05 per-unit value. The need for sensitivity in ground-overvoltage relaying makes accurate determination of ‘neutral-shift’ voltage quantities important, and otherwise not “buried in the mud.” [Reference IEEE Std C37.99-2000.]