I eliminated the interstage tranny in my AF-67 by converting the parelled connections of the existing 12AU7 into a phase inverter with 3 caps and about 6 resistors. My freq response was better. Looks like 70-5Khz on the modulated carrier.
I modified a ranger years ago by taking the DC off the primary of the driver xfmr, and then cross-laminating the transformer.
To take the DC off the winding, run a plate resistor ( as I recall, 30K but it may have been 10K - this was 30 years ago) from the parallel 12AU7 plates to the HIGH VOLTAGE +600 plate supply. Insert a 1 mfd 600v cap between 12AU7 plates and the driver xfmr primary. Disconnect the bottom side of the primary from the +250 volt supply and connect it to ground.
Fire up the Ranger. The voltage drop in the plate resistor should be just enough to maintain the normal voltage on the 12AU7 plates. (If it is way too low, try a 10K resistor instead of 30k). BTW, as I recall, I used a 5-watt resistor for the plate resistor. To be sure, calculate using Ohm's law. The audio should work as normal, with noticeable improvement in the low frequency response, due to the DC bias current being taken off the primary and elimination of the saturation of the core.
For much improved audio response, remove the driver transformer. Pry off the metal bracket that holds the laminations together and serves as a mounting bracket for the transfromer. Carefully remove the "I" laminations, then gently push the "E" laminations out of the core. It may help to warm the transformer in an oven to about 150 degrees F to soften the wax - but don't overheat it beyond that temperature, to the point of causing the paper to deteriorate. Now, reassemble the laminations into the core, alternating the direction of each set of E and I laminations, stacking them like a power transformer. The last set may be difficult to insert, but with a little care and patience it can be done. I would advise the following procedure: note the direction the last set needs to go, remove a couple of the other sets, insert the last set in the proper direction, then wedge the two sets you removed back into the core between the other laminations, and carefully force them all the way until they are even with the rest of the laminations. That way you don't risk gouging the paper insulating spool that holds the wire on the windings, as you insert the final set of laminations.
I recall that with mine, after cross-laminating the transformer, the frequency response was flat down to about 30~ after replacing some of the interstage coupling caps with higher values. I also removed the shunt capacitors that I recall going across the driver and modulation transformers. This immediately resulted in self-oscillation due to phase shifts that occurred at higher audio frequencies (most likely in the "ham radio quality" driver and modulation transformers), turning the negative feedback into positive feedback. I simply disabled the neg feedback loop by disconnecting the feedback winding on the mod xfmr, and bridging across the open connection in the speech amplifier circuit where the winding went.
That greatly improved the high frequency response, and I found that with the xfmr modification and larger coupling capacitors, the negative feedback was not necessary for good frequency response and low distortion. The overall freq response and distortion level were much improved over the stock circuit, and it required only two new components - the 1.0 mfd coupling cap, and 30k (or 10k?) 5-watt plate resistor.
One note of caution: if the Ranger modulator is to be used as a driver stage for a high power class B or AB2 modulator, for example for the desk kw, the feedback circuit is essential to make it work satisfactorily as a class-B driver, so I would leave the feedback circuit intact, and experiment with how little shunt capacitance can be used before self-oscillation sets in. I suspect that's the real reason for the neg feedback circuit, since the Ranger was designed to also serve as a driver for the Johnson KW.