My instinct is that, in terms of reproduction, it's appropriate to present what's on the recording (high fidelity with respect to the input). Otherwise it's just tone control - if implemented in hardware (e.g. crossover design) then it's applied across the board to recordings good and bad, making both (again, with respect to fidelity to the source material) inaccurate.

My assumption is that the recording already incorporates the tonal bias of the producer and engineer - am I way off base here?