Are you referring to psychoacoustic effects? Something like perceived loudness is only determined by waveform amplitude, but also affected by frequencies present? Or maybe the vocal and bass parts have transients at different times causing the other to be attenuated when it was already relatively quiet?

Other than that I'm not sure what else you could mean. Maybe we've been using different compressors. I've used a small handful of hardware and software compressors and haven't found them to sensitive to spectral content when used "normally". Meaning no extra filters or side chain configuration.