I'm now runnng the timer2 at 1:4, and that gives better resolution, but less delay time (abt 1ms I think) and this makes the retard angle quite low on low rpm's. It would be better to use timer1 I think..

I'm now running the trigger at abt 48degrees to give some time for the processor on high revs. I have changed the excel-sheet a bit for this, to compensate for the 1:4 timer frequency and pre-trigger. Again, this could be improved with a 16bit timer..

I also added a delay-compensation into the excel sheet. This allows me to enter the total delay time (of the processor, hall sensor etc), and that is then compensated in the actual values written in tspark.

Since the timer is now the limiting factor, I see values above 255 and sometimes below 0 in tspark. As it is now, I have to manually correct them so they stay between 0 and 255. It would be nice to incorporate this in the code, too much work to do it manually..
 I also tried adding a calculation so the actual angle shown in the graph is based on the corrected tspark value.

I'm not sure if I should continue with this concept, or hope that my friend can solve it using the 16bit timer1.
 
Practical tests:

The delay of the processor does not seem to be affected from changing the timer frequency, It's still around 60us.

When I set the delay compensation to 100us and a constant ignition angle, I get the same behaviour with the ignition timing lamp (strobe) as running without the cotroller board.
The behaviour is that I see a slight retard of the ignition angle at high revs (a few degrees) with and without the controller board. The tricky part is that I don't know if this is actual delay, or just due to a delay in the timing lamp..?
There must be some microsecond delay in the timing lamp too, but I'm not sure how much, and I can't think of any way to measure it.. Suggestions?