I have managed to get what seems to be acceptable range test results.

The main change I've made to to move to 72 MHz from 50 MHz. Before I chose the 72 MHz crystal, I scanned the ECU, looking for the frequencies with the least signal strength. I found quiet spots near the channels that end in '5'. E.g. channels 15, 25, 35.

Compared to the signal levels my scanner sees on 50 MHz, the 72 MHz band is much quieter, except for a few spot frequencies, like channel 1 which is only 10 kHz away from the third harmonic of the ECU's main clock. The second harmonic at 48 MHz is also there, but not as strong as the third. This is expected since pure square waves consist only of odd numbered harmonics.

Right now, I think the only way I'd get satisfactory results on 50 MHz would be to improve the RF suppression in the ECU. I'll stick with 72 Mhz for the moment.