The problem lies in a potential short to ground.
What are your RF power levels showing? Low power - or Rises then "holds".
Might want to "tap" the 8V regulated power supply. Monitor it from several taps - see where the "sag" begins - even the TX side back towards the Pre-Driver and monitor there too.
There may be a problem with the 4558 Mic Amp - TX / RX switch IC - IC 3 - although unlikely because you are able to even key the radio - it's when it runs a high current demand mode like SSB and even with voice peaks - which shouldn't make it act this way - it does this "thing" - so look at the main Regulation output at the Regulators own (8V constant) output - and while you can do this - check the TX voltage (TX Switch supply) and see if it' "drops out" as you continue this test
- - if so - you may have problems of it needing a good tune up and possible cap recheck and replace - for the TX 8V should remain steady during the keyup and even when your yammering and yelling into the mic. Voltage sags are a chronic problem with these radios usually caused by bad soldering; from simple cold solder onto the test point jumper mirror board can short to the jumpers UNDERNEATH and generate this condition - reflow solder around this section and look for potential shorts and even pockmarks from arcing
Although SMD - good example of what to look for
Can you adjust ALC and AMC and does the power level show any changes - AM/SSB?
Once AMC an ALC can be ruled out as part of the problem, the condition is with the "RF amp" or TX Strip or RF Deck - depending on your interpretation of the function.
How? Several ways, one mostly being bad coupling between stages.
Since this is affecting the power supply - to blame the power supply - prove it - simply remove the jumper and install instead, a jumper wire clip to jumper from the AM Regulator post to the Driver or Final test point post - using a Dummy load on the Antenna jack, to isolate the problem either as a bad Transistor (could be an MOSFET installed in error) or bad soldering onto open trimmer pot that lets all the bias into the Transistor it's designed to trim.
You're getting 50mA and it's adjustable. So - by your word, this rules out the MOSFET and trimmer problem - so we can only presume to continue with the above steps to rule out one section or another - if the lights "continue to dim" even while jumpering one section versus both from the same (the Mirror board test strip jumper) - but yet it (Dimming wacky performance) doesn't occur when NOT jumpered or even when the other section is jumpered nothing happens - during this test, the bad section is; the one you're jumpering causing the lights to dim.