Not sure about your method.
- I really wouldn't want to use an FG504 for adjusting the step response of a 7A26, be it for the slope or the post-slope time. I don't know what Tek 7000-model you have; did you try observing its calibration waveform and does it look identical to the FG504's at attenuation settings above 10 mV/div?
- I wouldn't call what I see straight overshoot. There seems to be some overshoot but your time scale is too slow to really see it well. After the overshoot, which itself seems to contain some fast "ripples", the level dips to almost the flat level but rises again not too steeply to a bump and then "slowly" falls off. I see that in your 7A24 pictures as well, also at larger input voltages (and attenuations).
- I guess you compared with another 7A26. "Consistent" misadjustment could explain the consistent behavior of both channels in your problem module. If the wave shape with another 7A26 and the same 7A24 setup is ok, the following doubts about the FG504's behavior don't apply of course.
- Things may look OK driving the 7A24 but without a 50 Ohm feed-through terminating the signal into your amp who knows what shape the signal really is, so what does the waveform via the 7A26 mean in that case? It looks like you did use a 50 Ohm feed-through on the 7A26 for all your photographs (on-screen amplitudes are equal) and that's ok but it could be misleading:
At 10 mV/div with the feed-through, the input voltage follower circuits are fed by a low impedance (your feed-through). In that situation, C130, C133 and C134 have relatively little influence.
At all greater attenuations, the series resistors in the attenuators make for a much larger impedance driving those voltage followers and the influence of C130, C133 and C134 is much larger.
If you didn't yet, have a look at the waveform on the 10 mV setting using a well-compensated (!) 1:10 10 MOhm input resistance probe. Obviously, you'd need a larger input voltage.
If the shape differs a lot from the 50 Ohm feed-through situation, I'd suspect a problem in the voltage follower circuits or an earlier misadjustment of those circuits in both channels, explaining why they are both wrong to the same amount. Again: C130, C133 and C134 and components around them (but these C's are the only adjustable components in the area).
The failure mode of the nylon-topped attenuation blocks leads to jittery behavior and finally, complete loss of adjustment influence by the problem-block; internal solder connections are bad and contact is lost.