The final system to model is an interface made of 2 surfaces of 12 fcc-Pt atoms each, with a NH3 molecule in between. I'm using ultra-soft pseudopotentials. If I take the energy cutoff (ecutwfc) that gives a converged lattice parameter within 0.01 Angstrom, I find 30 Rydberg for Pt and 40 for NH3 (NH bond length converged within 0.01 Angstrom).
My question is, given a large number of atoms with lower cutoff and a minority number of atoms with larger cutoff, how should I pick the final cutoff? Should I always go for safer option (higher cutoff) even though it causes the majority of the system to be treated with a cutoff higher than needed, increasing computational time?