Hi Fred -- The audio is in fact taken from Z4 as you say. But the energy appearing in the secondary of Z4 is a modulated 455 kHz envelop, with the audio modulation appearing on both the positive and negative excursions of this 455 kHz envelop "carrier" (it takes an audio signal representing 50% of the RF Carrier Wave signal to modulate the carrier 100%. Because it is modulating both sides of the carrier, a 50% audio signal effectively then causes the carrier wave to swing between a value of 0, and 2X the carrier signal alone at 100% modulation). Since the modulation signal appearing then on the positive excursion of this carrier is 180 degrees out of phase with the modulation signal appearing on the negative excursion, they cancel each other out, so that if applied to a speaker, no audio would be heard -- unless -- one side of the envelop is "chopped off". That would allow just one side of the modulated envelop to appear at the output of Z4, with audio that could then be heard. That's the job of the detector stage: by connecting one side of the Z4 secondary to ground through a diode (in this case, the cathode/grid elements of V4), and the other side of the secondary to ground through a capacitance that only reacts to the 455 kHz energy (C58), one side of the envelope is then effectively ignored when the carrier swings in the direction that reverse biases the diode elements of V4, relative to the charge on C58. When the carrier swings in the other direction, V4 conducts, completing the circuit, so that the audio modulation on that side of the envelop -- and only the modulation on that side -- is then "recovered", and heard as the original audio modulation signal. Think of it as an audio signal that is recovered at a 455 kHz sampling rate.
The detector diode could be connected either way, to recover either side of the 455 kHz modulated envelop. But it is most often (and conveniently) connected with its cathode end grounded. This then causes the voltage appearing at terminal 2 of Z4 to not only contain the recovered audio, but also a negative DC voltage that is proportional to the strength of 455 kHz energy appearing in the secondary of Z4. This negative voltage is often used to provide AVC bias voltage (Automatic Volume Control) back to the RF and Converter stages in the front end to help prevent "blasting" on strong signals. In this case however, the AVC voltage is developed elsewhere in the circuit (by CR1), so that the negative voltage at terminal 2 of Z4 is only used to drive the tuning indicator (M1 in this case) to indicate relative signal strength. Because the meter is a DC device, it responds only to the negative DC voltage developed at terminal 2, rather than the AC voltage representing the audio modulation. Therefore, it's indications are almost entirely independent of the audio modulation present, and indicate only the DC component, representing the strength of the 455 kHz carrier -- itself, a product of the strength of the RF signal received at the antenna.
In this way then, V4 is a vital component of the AM tuner in an 800C or 1800 receiver. Yank the tube, and the AM section will go dead as then, the secondary circuit of Z4 is permanently opened. The fact that the secondary of Z3 is in series with the secondary of Z4 is insignificant. Because the secondary of Z3 is tuned to 10.7 mHz, it appears as a dead short to a 455 kHz signal, so that its presence in the circuit when operating in AM mode is invisible.
I hope this helps!
Dave