My preference would be listening very closely and analytically to those ref tracks to work out what's happening in terms of volume (individual instrument levels, relative levels and overall levels), and how the instruments work together to complement each other (eg musically, intensity-wise, pitch-wise, tonally, spatially, mix-wise and mood-wise). Although the ref tracks might be virtual they will be designed to emulate how real players work, which is not just simply playing at the same time as each other, but to playing together in synergy and create a musical journey that evokes the desired emotional response, and even FX like reverb and delay are there to enhance the sonic texture. It might help to make notes on what each instrument does and how the parts 'dance' around each other and respect each other's space.
Ultimately, you could argue that making cues is fulfilling a functional need rather than creating high-end art, but the cues that get placements are chosen because they support the mood of the screen action, and its our job to provide music which does that..... and get the waveforms right
.
