consider that there is a mechanical latency in the pulse width measurement. 800us is not double the fuel of 400us. The latency measurement appears to shorten with increasing pressure, as if this problem wasn't complicated enough.

FWIW, I typically have not increased pressure, in fact, on some platforms I lower pressure at lower loads. So I use a wider range of pressure, spread across a range of loads. I do this with a timing adjustment that allows for greater economy, and lower noise.

Your post is a good discussion starter