When you offset a DBM it becomes temperature sensitive so carrier level will change with diode temperature. It also reduces the dynamic range of the mixer. So just use it to generate DSB then add a combiner after it to inject the carrier. A step attenuator would work fine. Then you just need a splitter to provide carrier to the DBM LO and the carrier inject path. The DC offset method will work fine though.
With the upside down tube circuit, the carrier level is rock stable, since you set it by adjusting DC plate voltage. Zero DC voltage, you have suppressed carrier down to -30 dB or so. Full plate voltage, you have regular plate modulated AM. Anywhere in between, you have a reduced carrier which is a function of the DC plate voltage.
The screen modulated balanced modulator like the DSB100 has considerable distortion, since screen grid modulation is inherently non-linear, especially at the extremes of the SG voltage swing. With regular AM, you can minimise distortion by limiting modulation to 85% or so. But with the DSB100 circuit, you are driving the modulation of each tube beyond 100% in both directions.
I have a little UTC LS series modulation transformer rated for about 20 watts of audio. I could use that to modulate a small class-C upside-down tube final, then amplify that with a big leen-yar. Or I could use the modified Eico 730 modulator that is presently in my 10m rig.
I remember the old phasing type slopbucket exciters back in the early 60's, like the Central Electronics series, Heapshit SB-10 and Johnson Pacemaker, in which the carrier null drifted badly and had to be fiddled with constantly during the course of a QSO.