Well, first you can eliminate any TR3 units. They are smaller and don't put out as much power as the ones used on later TR3A through 4A.
Next, I would pull them apart for a visual inspection. Any signs of overheating, like thrown solder, or burned-looking armature windings, or insulation flaking off the armature windings would be an immediate disqualification. Units with these defects will sometimes pass tests, but not work properly on the car. Also check the tape on the field windings for deterioration. The commutator should be no darker than a medium brown; black generally indicates a problem. Brushes should move easily in their holders, minimum length is given as 11/32". Any sign of rubbing between the armature and pole pieces indicate the rear bushing is worn out (or the front bearing is failed). The front bearing should turn silky smooth, any roughness is grounds for replacement. (Luckily they aren't expensive and are easy to change.) You can judge the wear in the bushing by putting the rear cover on the shaft and rocking it back and forth. Should be only a bit of play, not much.
Check the resistance from the field terminal to the case. It should be close to 6 ohms. Less than 5 indicates a shorted winding. (Again, the unit can appear to work with a shorted winding, but will quickly overheat in service and/or burn up the control box. I burned my thumb badly last year from touching the regulator relay after I failed to notice a shorted field winding.) Check the resistance from any commutator segment to the shaft, it should be infinite.
Reassemble the unit and check resistance from the output terminal to the case while slowly turning the armature. It should stay very close to zero. If it jumps around, you've probably got a bad winding. (Could be bad brush or commutator segment, but you already checked those.)
Beyond that, I would suggest mounting on the engine, polarizing, and running through the "on-car" tests.
https://drive.google.com/open?id=0B2H2NJt34OffNTc3ODkwYzAtYjRlYS00NDNmLWI0YTYtNjY5ZjQxZTA2NGFm
You really can't test on the bench without a lot of special tools.
No doubt someone will trot out the old "runs like a motor" test, but I've had quite a few generators that would run, but still not do their jobs. IMO it's more of a parlor trick than a valuable test.