QTX 475 kHz

output V over 50Ohm  ~72W Rf   at input  76W (25Vx 3A)   N=96% ! ?

V drain puls

had to modify the total capacity  of C2+C3 to 10+6.8+6.8 nF
and the output C4 6n8 + 6n8  to archive maximum output.
But at 25V still could not produce 100W out
I did change R4 to 1k  because of latching problems I had earlier that cost me a few IRF's 
Drive level increase  does not change the output in this class D  type switching mode.
The Flex1500 XVTR/C port  0dBm at level 30% is sufficient.
Also the K3 XVTR out level was perfect to drive the QTX. 
Class D is perfect for CW WSPR modes.

The Drain pulse form seems correct.
Optimising L3 for maximum output did not bring  more output. Po drops as soon as a ferrite bar is approaching L3.
I would expect that L3 could be increased when dropping the frequency from 500 to 475 kHz

After re-arranging  C2 C3 C4
The output increased to 75V/50Ohm 112.5W
50Ohm dummy Bird 8251

C2+C3  is now 10n+ 6n8+ 4n7
The C4 original was 6n8  now increased to 9.9n

At  DC input power I=5.23A at 23.5V =122.2W  n=92%

There is  a bit more heat developing in L3  the IRF540 is a bit warm.
The high efficiency is amazing.

The SWR power meter now points at 45 never reached that level before
it was only "12" in 2010 when WD2XSH/17 2-way WSPR succeeded