[Amps] Johnson Thunderbolt questions

[R.Measures]

>I'm in the processing of bringing a Johnson Thunderbolt amp up and can't
>quite make sense out of what I'm seeing.  I'm hoping there is someone out
>there with experience using the Thunderbolt (or a pair of 4-400A's) as an AM
>linear.  I've searched the list archives and found lots of old threads about
>Thunderbolts, but none of them addressed my situation.
>
>My Thunderbolt appears to be a completely stock, factory-built unit (with
>the exception of a 120 ohm resistor in parallel with the original 350 ohm
>swamping resistor).  The HV power supply is putting out 2200V to the plates.
>I've been able to get a good match to my IC-730 by using a small matchbox.
>I've also driven it with a Ranger.  The additional swamping resistor lowers
>the input impedance to about 89 ohms.  The 40 watts out from the Ranger
>gives 59V RMS at 89 ohms which is what 10 watts will yield at 350 ohms (the
>original AM/SSB drive spec).  Since the grids of the 4-400A's are biased
>at -75 volts in linear mode (Class AB1/2),  that sounds just about right.
>
>In CW mode (class C) I have about -165V on the grid, and I can tune it up to
>1000 Watts input, which is giving about 600 watts out.  So far so good.
>When I switch to linear mode, I can't get more than about 125 watts of
>carrier out (500 W PEP on voice peaks) before the waveform starts to flat
>top.  I can tune it up per the manual instructions and get about 350 watts
>of carrier out (360mA plate current), but there is very little headroom so
>I'm still only getting 500W PEP.  The original specs for the Thunderbolt are
>800W input on AM.  Assuming 35% efficiency for AB1, I'd expect that I sould
>be able to get about 280 watts of carrier and about 1000-1100 watts PEP.
>I've tried using it on SSB and I see the same 500W PEP output limit.  I've
>also noticed, that when operating this way the plates start to glow red
>after about 20-30 seconds of AM carrier (the plates never glow in CW mode)