Thanks, Henrik.
I had used similar code in the past with PBP2.x, but not identical, the crucial difference was that the prescaler of my earlier code was divide by 256. So didn't run out of ISR processing time before the interrupt was flagged again.
Realised that the ISR needed a few clock cycles, but did not realise that it needed so many that I would run out of time before the ISR was finished. I thought the ISR handling time would just extend the period of the timer by up to a microsecond or so. Now a little wiser!

As for my 0.75 us instead of 750 kHz (equivalent to 1.33 us), I should know better than to attempt simple mental arithmetic instead of pen and paper or a calculator!
Pinsamt, as they say in your neck of the woods.

Will go to 20 MHz crystal and a higher prescaler setting, to see how low a timer period is possible before the problem rearises.

Norbert