I've been chasing this HT related problem for the last 9 months. The symptoms are very straight forward: when cold and using a small Lucas coil, I have a good 15mm spark from any one of the plug leads to earth, but when hot (i.e. after 10 miles), the spark is barely visible and sometimes I limp home on 3 cylinders.
I did all the usually things...and more:
Tried 3 different condensers
Tried 3 different coils (coil ignition on the Hawk)
Changed the plug leads
Changed the base plate in the distributor
Changed the points
Added an additional earth lead from the distributor base to frame
Changed the rotor arm
I've now largely cured it; it was the distributor cap. This may seem an obvious thing to check but this cap is a new, modern reproduction one; and I have a couple of these and they all suffer the same problem; when hot, they fail...and it's not due to tracking etc. The final proof came when someone lent me their original Lucas bakelite one from their 1930's Austin Seven; this has the same cap as the Hawk.
Talking to Tim Green at Green Spark Plug Company, he suggested that modern plastic composites are fine when cold but if they get hot then.... and suggested I look out for a NOS bakelite cap. I now have one and I'm back on the road in a more reliable fashion but on to my question.
During all of this testing, I strapped this (below) 6v Chinese coil to a down tube and I'm still running the bike on this. The spark from this coil to earth is about 40mm (I kid you not!
), but even with this good bakelite distributor, the spark from a plug lead to earth is reduced to 20mm when hot. Obviously, a 20mm spark is going to be fine at the plugs, but why cant I get the same size spark out of a distributor as the spark that goes into a distributor? Where is the missing 20mm going 
?BTW: this little £23 coil is great! And at 50mm at its widest point, I should be able to hide it in the shell of the old Hawk coil which is housed in the dynamo body.
