My speculation is that we may both be from Missouri, the "Show Me" state (except I recall that you now live in GB?)
Absolutely NOT, I'm English born, breed and resident.
the iSis measured profile for HP Everyday letter size 8.5x11" glossy gives mediocre results, for me, to use with 5x7" sheets. To me, that suggests that 5x7 HP Everyday Glossy is different from 8.5x11" HP Everyday Glossy. Yes, certainly that could be lack of experience, care in "experimental design", and/or ability on my part.
Do you seriously think that HP (or any other manufacturer) would use different construction processes for different paper sizes of the same paper type ?
Have you ever come across
any manufacturer offering different profiles for different page sizes of the same paper type ?
There will be differences in how images appear dependant on their size, but that's a different issue to the pure colour management of the printing system.
Why not do tests to check ? print out a chart at the same size on different paper sizes (cut down to size if you want to check larger sheets) then measure them and see if there are differences that can't be explained by the repeatability of your printer and measurement system.
but "close as in grenades" rather than "close as in De2k < 0.5"
First of all what De2K are you discussing about ? Average of all reading? best 10%? worst 10%?
So the issues you need to consider are:
How repeatable are the reading of the same chart ? read the same chart(s) often and at different times. Do you always get exactly the same reading ? I'll bet you don't.
Does a smaller patch size (as ACPU delivers on Windows) effect measurement accuracy ?
How consistent is your printer ? Print the same chart out several times, then read it, how much does that vary ?
Once you know the above, then measure the difference between CS4 output and ACPU output. Is it significant ?
I've done all this. First when ACPU was released with an i1Pro and again with a newly calibrated DTP-70.
The DTP-70 is more consistent than the i1Pro.
With the DTP-70 Average read accuracy on a standard 6mm patch is 0.44 De2k, that falls slightly with the smaller 5.76mm patch from ACPU to 0.61 De2k.
Page to page consistency is around 0.58 De2k on my Epson Pro 3800
Average measured differences page to page between ACPU output and CS4 output is 1De2k, take away the read error and printer error and that's not a lot.
There are other sources of error that may influence results, maybe temperature of the spectro makes a difference and it's certainly very important to let the chart fully dry before reading (different papers and inks will behave in different ways).
Overall I'm happy that the output from ACPU is so close to CS4 it's not significant enough to worry about in real world use.