I did a test today trying to print a small test target without color management on the Pro-1000. Just 63 patches with a good spread around the gamut of Canon Glossy II paper.
...
Purpose was to see if different techniques of printing target without color management (all seen recommended here or there, as you would assume they were identical or pretty close) are consistent in behaviour, and if different, by how much (insignificant or major discrepancy). That has been done already above, showing major discrepancies using the software and setting combination as described.
Such a target is used as base to create a profile (how does device input RGB actually render in perceptual/tristimulis vision we agree on, e.g. CIE Lab, and reverse), with multiple features, one being 1) consistency between soft proof on monitor and what you get out of the printer, another 2) consistency with other printers, so you can trust that colors within common gamut abilities actually can render the same. So if multiple people individually and correctly profile their printer and print the same image on same paper with colorimetric rendering intent, they should look the same - within common gamut. If printers were even the same, they could share the very same profile, given printers were calibrated (as for instance the Pro-1000 can be).
So I currently wonder what will happen if you profile with two significantly different looking targets, like is currently possible, unknowingly, because PSP differs vastly from ACPU/ColorSync. Would 1) be achieved, but 2) not? That's a guess currently, but not thought complete through. Comments please. But I find it hard to believe the different target prints won't result in different profiles, making at least 2) impossible to achieve. And I would aim for having both 1) and 2) right, so we each agree on colors, getting the same consistent print output, and not just are satisfied that our own soft proof looks like our own print.
So if the target print does matter and causes inconsistency if not done right, how do we determine "right"?
Well, I thought that if Canon has gotten their own profiles roughly right, then I can predict using that OEM profile if my target prints look "right" (as their profile includes the device RGB input mapping to Lab). Having made my small test target preconditioned with Canon's OEM profile, I have, using ArgyllCMS, included the expected raw print output for each patch RGB device input - as Canon have measured it when creating their own profile for a similar combination of printer/paper.
So, does the PSP print match the expected values from Canon's own profile measurements?
colverify target1-case-2-PSP.ti2 target1-case-2-PSP.cie.ti3
Verify results:
Total errors: peak = 58.401719, avg = 14.152884
Worst 10% errors: peak = 58.401719, avg = 43.449070
Best 90% errors: peak = 32.835848, avg = 11.069075
avg err X 0.063412, Y 0.058000, Z 0.021320
avg err L* 6.300949, a* 7.518214, b* 8.974196
No, not even close. With peak dE of 50+ and worst 10% error averaged of 40+, I will surely conclude a no. The print does not in any way represent the expected raw device response.
Does the ACPU and ColorSync target prints (which are similar to each other as shown earlier) match the expected values from Canon's own profile measurements?
colverify target1-case-1-ACPU.ti2 target1-case-1-ACPU.cie.ti3
Verify results:
Total errors: peak = 6.774103, avg = 2.858761
Worst 10% errors: peak = 6.774103, avg = 5.876139
Best 90% errors: peak = 4.843732, avg = 2.541143
avg err X 0.010225, Y 0.009825, Z 0.008567
avg err L* 1.287127, a* 1.323744, b* 1.461922
Yes, to a degree that could reasonably be explained by printer, paper, environmental and measuring variations, and an indicator for why you might want to profile on your own as differences can show up. Peak dE of 7, worst 10% average of 6 and best 90% avg of <3 (under 1.6 in CIEDE2000) doesn't sound unrealistic to what I would call an expected match - certainly compared with the PSP results.
So I conclude that ACPU/ColorSync are the ones to trust for target prints, both agreeing with each other and Canon's own measured device responses contained in their OEM profile.
Please let me know if I got this line of reasoning all wrong, or someone agrees.
If you agree, I will hope PSP/macOS/driver gets the printing with "No color correction" correct, as it is useful in several scenarios, if done right.