1) Both designs pull the test voltage up to +V. That will activate a PN junction. Either design has a threshold below this level, so their accuracy isn't affected. But the voltage is there.
2) Essentially correct. The unadjusted input offset voltage of the first 741 is nominal 3mV, which makes its "zero" output +/- 50mV. It either requires adding an offset null pot on the 741 or being very lucky.
3) The second 741 should be equally unpredictable. The bias current through 100K helps, but not enough to compensate for up to 3mV of offset voltage.