Thanks very much for all the feedback I've received, both on and off the list. 
Jim seems to pretty much cover all my points, thanks. The only minor followup 
question I have, that might be of general interest to the group, concerns 
loading coils

> if the horizontal section is much longer than the vertical section,
> the vertical current distribution will be nearly uniform, less current will
> be capacitively coupled into the tower, and the effect of the tower and/or
> loading coil will be minimal. If the horizontal top-loading section is
> shorter, adding an elevated loading coil as you describe will have more
> beneficial effect. This rather depends on whether your inverted L is cut for
> operation on the 160m band, or is 160m long. The effect also depends on the
> spacing between the downlead and the tower.

My 160M inverted-L is 15M vertical and 35M horizontal (with a series 1000pF 
variable capacitor at the base, it looks like 50+j0 ohms on 1830KHz). For 
mechanical reasons, I've compromised the downlead arrangement: the wire runs 
through a plastic pulley at the top of my tower and so is very close to the 
metalwork up there. At the base, I am using low grade insulation for the wire 
run into the ATU shed, because the voltages are low on 160M.

I fear the tower proximity will degrade performance, but I have to live with 
it. The horizontal top could be extended, perhaps from 35M to 50M, but not 
enough for resonance on 500KHz. Multiple loading wires are aesthetically 
unattractive. In addition to reducing tower coupling losses, the elevated 
loading coil looks attractive as a route to reducing high (reactive) voltages 
near the base; even 500v makes me nervous with dogs, wives, and low grade 
"160M" insulation on site!

Noone is offering me 500KHz crystals, so it looks like I'll be scratch building 
a transmitter using a mixer VXO or a ceramic resonator.

Thanks everyone,

Alan G3XAQ