Guilin Sun
Ansys Employee

I did not find description of collimated beam coupling to fiber in the paper. Usually the fiber is long enough to do two simulations:
1: coupling coefficient from the light source (mostly from a semiconductor laser or  a laser beam through a microcope object lens with numerical aperture NA).

2: the sensor, SPP part with analyte.

I would suggest you simulate the sensor first. You may also need to check if the structured SPR device creates the reflection in to the fundamental mode or with some higher order modes.