Steven,

Your question about "the best method" is hard because often what you think is the best may be in contradiction with what a Code imposes.
A good question is what B31Ea considered as appropriate in their 3.4 "Design by analysis".
What they said in 3.1 is that "the seismic loading shall be specified for each of three orthogonal directions" combined by SRSS or using an alternative method (envelope of the SRSS E-W+Vertical and N-S+Vertical). There is an warning about including in-structure amplification of the free- field acceleration by the structure as seismic loading applied to piping systems and no clue about SAM. At least it seems clear that SRSS is applied for effects of "seismic loading" in three orthogonal directions- to comment your proposals.
Without B31Ea support for "seismic loading" applied in one direction, there is the temptation to absolutely combine the inertial response and the pseudostatic response (peak group displacements occurring at the same moment, combined in the most unfavorable way to calculate SAM component of the seismic response). The problem is this approach "can considerably overestimate the seismic response when compared to time- history".
NUREG-1061, volume 4 (reference https://www.nrc.gov/docs/ML1134/ML11343A034.pdf ) gives you what they consider as better alternatives. You can find the document on internet and study.

BR,
M