I'm no optics expert, but I guess an integrating sphere might be used if the amount of reflectance was likely to change. My guess is that provided your incident angle is 90deg and the change to RI and absorbance is not enormous, you should be ok.

An integrating sphere is the gold standard for quantum efficiency measurements. The reason is the uniform Lambertian reflectance of the inner walls of the sphere. Using an integrating sphere is better than just 'taking the spectrum' as you mentioned above because the integrating sphere collects all the light emitted and has uniform response no matter the angle of the rays. Whereas, for a typical photoluminescence measurement you would use an objective lens with finite numerical aperture and uncalibrated transmission of different angles. Note that PL and EL are typically highly diffuse in the angular distribution.

As I understand, you want to know while a certain detector in a certain arrangement can measure the pure laser and combination of the laser+luminance powers, why should we use integrating sphere! Note that it is a little simplistic to think that an optical detector correctly shows the power it has achieved in any way. In fact, in a detailed view, the response of a detector is not quite uniform all over its sensitive area and, on other hand, it depends on different parameters of the incident light such as angle of incidence and polarization that in turn, may complicatedly vary within the incident beam cross-section. The performance of integrating spheres effectively eliminates all these dependencies and provides similar condition for all the measurements. Direct measurement of the light power (without using the integrating sphere), of course, is still reasonably applicable in most of the ordinary measurements, but not for most of the sensitive luminance related measurements.