Thank you everyone for your helpful comments. I have done some more tests on my 
inverted L system with the following results (please note that E1 means a 
measurement made with the first earth connection, E2 the second and so on):

Feed-point resistance at 137kHz in ohms: E1 = 156, E2 = 156, E3 = 200, all in 
parallel = 135 (this figure is about 15 ohms less than the value obtained last 
month).

Individual earth resistances: E1 = 41, E2 = 41, E3 = 85  (in each case in 
series with 7 ohms inductive), all in parallel = 17.  These measurements were 
made by taking 2 earth connections at a time, measuring the loop impedance and 
estimating the individual components.

Taking the differences: E1 = 115, E2 = 115, E3 = 115, all in parallel = 118. It 
looks as if this very nearly constant figure represents the resistances of the 
return paths from the remote ends of the horizontal section to the nearest 
earth spikes.  This means there is nothing to be gained by adding more earth 
spikes.

Deploying an 8 by 4 metre loop on the ground below one end of the flat top, and 
connected to the nearest earth spike, reduced the 135 figure above to 126. 

Measurements were made with a GR1606A bridge.  The generator and detector were 
both isolated from the instrument with double-wound transformers. The case of 
the instrument was connected to the earth system under test in accordance with 
the manufacturer's instructions.  Note that the earth resistance measurements 
were the same whichever way round the earth connections were).

Note also that the transmitter system is isolated from the antenna by a 
double-wound transformer, so as to isolate the RF side of the station from the 
domestic electricity earth.

Tomorrow's step is to put a loop at the other end of the horizontal section and 
see what effect it produces.

John F5VLF