A simulation of the circuit with a shortened duty cycle did show improvement in the waveforms and efficiency but this may not be the best approach.
Following Ermi's comments about tuning the amplifier to the load I changed the value of C1 and this had a marked effect on the waveforms and the efficiency.
As time permits I will try this tuning change in the real transmitter and will report back.
This shows the value of peer review where others check and critique designs which can promote progress. We'll see how well this works in this case.
Neil
50% duty cycle is typically used for class E, but a wide range of other duty cycles will also work. The published formulas are for 50% because the mathematics is simple enough for analytic results to be possible. Numerical analysis is typically needed for other duty cycles. A duty cycle of 33.3% will maximize the operating frequency in case high frequency operation is needed.
Some tuning to obtain class E conditions can also be done by varying duty cycle. A circuit that is operating near class E can sometimes be accurately tuned in by slightly raising or lowering the duty cycle.
The ability to fine-tune the duty cycle, since it's so critical, sounds like a nice feature to have.
Would there be a way to meter the duty status so the un-sophisticated user could get it setup?
With C1 (specd. as 1000pF) varied from 600pF to 1330 pF the maximum efficiency was measured as 86.2% with the original design value of 1nF so it appears that this is optimum for the R load of 29 ohms.
A check of the collector to ground voltage with a 20 MHz scope revealed this waveform: 
This looks better than the simulated Vce. Ermi, what do you think?
It would be more telling if I can scope the collector current but this requires major mechanical disassembly and I skipped this for the time being.
Neil
Carl,
The efficiency was not very sensitive to the duty cycle in the bench test. The simulation seems to exaggerate this effect. I don't think it is worth worrying about.
Neil
Neil,
Strictly speaking, there is not supposed to be any ringing or undershoot of the collector pulse, but I know that those glitches are very difficult to get rid of. I can see the resemblance to a class E collector pulse.
Carl,
I think that the easiest way to get variable duty cycle is to use a 74HC or HCT one-shot, like the 123 or the 221, to drive the MOSFET gate buffer circuit. A pulse width adjustment would be necessary. As for "metering" of the duty cycle, an oscilloscope would be needed.
Thanks to Ermi, RFB, Carl, Bruce, Phil, Robert, and others who have posted questions and comments about this design. Such discourse is very helpful to me.
Phil has suggested that the output filter be eliminated to raise the efficiency with the selectivity of the coil loaded antenna system used to suppress harmonics. This is a good suggestion for anyone who decides to build and experiment with this transmitter however I am not going to pursue this since the unit as built works very well.
Ermi,
The ringing was observed previously when I was prototyping this and similar circuits and I think it may be due to higher order effects caused by the Ls and Cs in the output filter. Fortunately, this does not get through the filter to the output. Perhaps it is caused by circuit layout but I did include wide conductors and an other side ground plane in the construction on the PC board though at this frequency this should not be critical.
Neil
"Phil has suggested that the output filter be eliminated to raise the efficiency with the selectivity of the coil loaded antenna system used to suppress harmonics."
I wonder what kind of effect that would have on the harmonics and spurious suppression without a filter when the inductive effects take hold on the outdoor antenna and loading system. The shifts caused by the inductive reactance of external influences de-tunes the system, thus will also de-tune the filtering characteristics trying to be achieved. Maybe not so much to worry about, but something to consider and perhaps take a look at.
RFB