As I understand the circuit a voltage of about 2.8 volts is derived from R1641 and R1636 two 2.2 megohm resistors in series.
The several modifiers of the voltage to reduce the voltage to the IF pull that voltage down when the op amps voltage drops from 7 volts and bias diodes D1606 - 1608.
I think that the FET Q1606 is controlled by the UNSB line from the logic CPU and this bypasses the network of R1626 and R1634 through D1610, if that makes sense. Q1606, when turned on connects R1627 in parallel with R 1641 and 1636.
When not turned on the diode D1610 provides the voltage at the junction of R1626 and R1634 at about 1/3 of the 8 volt buss or 3.1 volts.
Either way Q1606 controls the ALC base voltage and the amount of drop when current is drawn. The voltages are close either way on my working radio.
In my radio the ALC line is pulled hard down when UNSB goes low. Next I will try changing Q1606 as I think that it is letting the line which goes to UNSB pull the ALC voltage down because perhaps the FET is "Leaking" the ALC voltage to ground through this line.
And thoughts on this?
I have included a pdf of the main board ALC circuit for those who are masochists and want to follow along