I guess I assumed that if the bandswitch was set to 20M it would act like a bandpass filter and pass a 20M signal as well as the second harmonic of a 40M signal. Clearly my understanding is deficient.
There are only two tuned stages in the DX-60 from what I recall, the PA tank and the grid circuit after the 6CL6 multiplier. So, if you inject a 20 Meter signal into the grid of the osc. 6CL6, you will essentially have two 6CL6 buffer stages. I'd be very leary of applying a five watt signal into two low-level stages, you're going to drive the heck out of them. The best remedy(IMHO) would be to link couple directly to the PA grid tank circuit. That would involve extensive modifications to the DX-60. You could try using a power attenuator between the five watt exciter and the input to the oscillator, that will help with the excessive drive and gross impedance mismatches as well. Maybe going back to the pre-driver in the SB-104 might be a consideration? It won't be an easy mating in my opinion.
Pete