thank you for your fast and extensive reply. - I am all confused...
The thought to use PhotoZoom followed by FocusMagic was trying to combine 3 of your recommendations: PhotoZoom as the best tool for up-sampling to large prints; deconvolution sharpening AFTER up-sampling (but still before print sharpening); and FocusMagic as the best tool for this (or one of the 2 best, together with InFocus).
But if the PhotoZoom result is not recommendable for subsequent print sharpening, I will avoid it.
The question is whether subsequent sharpening after upsampling with Photozoom Pro will add much.
It depends on image content. Photozoom Pro is very good at two things. First it does an excellent upsampling without adding artifacts, and second it adds real edge resolution and also some beyond what's in the original image. Expecting it to allow additional
deconvolution sharpening may be too much to depend on. It might succeed, but it might also fail.
So that leaves up-sampling and subsequent capture sharpening. If PhotoZoom is to be avoided, I could use PhotoLine's Lanczos (3 or 8, btw?), followed by FocusMagic. Would these be the best tools available to normal mortals (outside PixInsight)?
The full workflow is; Capture sharpening (preferably by Deconvolution because that restores real resolution) and after some Creative adjustments and resampling for output, Output sharpening.When the resampling for output doesn't add sharpening by itself, Deconvolution output sharpening is advised.
However, Capture sharpening and Output sharpening can be combined in a single operation when the intermediate 'Creative' and upsampling steps do not
affect sharpening. When the intermediate 'Creative' steps do
affect sharpening, then one should use very good Capture sharpening, Creative steps, and Deconvolution sharpening after resampling.
I have never heard of a tool that could do capture+resampling deconvolution in one step. By "one-step" I only meant I would rather do 1 deconvolution at the end of the workflow rather then accumulated ones, not that it had to be combined with up-sampling.
(The workflow implies exposure stacking with ImageFuser followed by Focus-stacking with Helicon.)
Can you de-confuse me? :-)
Maybe it helps to follow the logical path that the image data does. Whenever a source of blur is introduced, one has the option to immediately remove it. But that requires optimal removal without introduction of artifacts that can wreak havoc later by accumulation errors upon errors. One often has the choice to combine several steps, e.g. cascaded Gaussian blurs, by a single blur and thus a deconvolution to reverse that blur. Operations that significantly change the image data, e.g. by adding artificial (but convincing) data such as edge resolution or noise (like Photozoom Pro does), will complicate the combination of multiple (prior and later) steps into a single one.