[permo-w]

I suspect that a more nuanced approach could suit general users the best. Retain boldness, italics and underlining, change to fit target colour and font.

Boldness, italics and underlining actually denote meaning, whereas font and colour are generally just aesthetic

[derefr]

In real Desktop Publishing software, text (content) and its style (layout) are separate — the text usually comes in from an external source (e.g. a file on an SMB share; a document in a CMS) and can be updated independently of the layout.

The linked text format is usually Rich Text (RTF). This allows a lot of things, but the Desktop Publishing tools only interpret the tags for bold, italics, underline, and a few other things (strikeout, subscript, etc.) All other styling in the linked text, they throw away.

This is precisely because, as you say, those specific styles actually denote meaning. They're something the writer adds. No other styling is used from the linked text, because none of the other styling is the writer's job.

All other styling is instead applied to the block(s) within the layout where the text gets embedded into. It's the layout designer that gets to decide the font, size, spacing, etc. for the text. Those attributes aren't stored with the text; they're stored with the layout.

To me, this makes far more sense as a workflow, even if you're a single author. I constantly wish that "word processors" had restructured and absorbed ideas from Desktop Publishing software when it came about. Instead, we got the garbled hybrid: you can have "document styles" like Title, Heading, Body, List Item, etc.; but they are essentially markup, moving around with the text (rather than there being any concept of an "section of the document" that gets styled, that text can be moved into/out of, and where the styles of that section will apply to the text only while it remains inside that section, such that moving text out of that area doesn't copy the styles of the section, only the styles of the text.)