The problem is in the transformers used in this bench version, or their application. We don't know what they are. That's the 'ideal' solution, be it finding a more optimal loading condition, replacement, etc. We don't know that the external parts have all been proofed for best practice yet. Frequently I see treble change drastically depending on loading. Actually, we don't know if the input attenuator is in use, its secondary load is present, how it behaves without it. How it behaves without the input transformer. That all seems right to me given the OP wants to experiment with various existing PP circuits. An EQ addition is the band-aid if all else is chased down and nothing is to be corrected. Yes, parallel cap resistor for a 1st order filter between stages, or 5dB NFB, see which sound is preferred. Loss of gain either way. Collins did a couple very low NFB tweaks in this era, so that wouldn't be out of context much.
Collins 6S is the console module version of this amp, what I have to look at. The attenuator is simplified to a 3 step, and the metering is changed to a series 31R5 wire wound in the B+ line, since the other console modules all measure from B+ across a resistance, measuring an entire module rather than individual tubes. Both cathode resistors join at B-.