RSI's maps are by their very nature non-essential, so I think it would make sense if it always deferred to other maps. I'd accept something like the following:

function! s:Map(mode, flags, lhs, rhs) abort
  if empty(mapcheck(a:lhs, a:mode))
    exe a:mode . 'map <expr>' a:flags a:lhs a:rhs
  endif
endfunction

While your suggestion would avoid overwriting existing mappings, I don't think it would fully resolve that problem of conflicting with readline.vim. That would only be possible if vim-rsi is loaded after readline.vim. Some plugin managers support such a thing, others don't. And when not using any plugin manager this is even harder to achieve.

Also it would not help if someone wanted to have readline-style keybindings only in command mode (like readline.vim does) or (less likely) the other way around.

Therefore I don't think I will make another PR with your suggestion. I can live with having my own fork of it.

Thank you for having a look into it!

As @justinmk explicitly "dislikes" this PR I am a bit curious about what he dislikes about it…

It will work regardless of load order assuming readline.vim doesn't also implement this behavior.

Ah, I see.
If vim-rsi is loaded first, it will map these key and those will be overwritten later by readline.vim. If readline.vim is loaded first, it will not overwrite these mappings.

Still, this doesn't allow for a user to activate vim-rsi only in insert mode or only in commandline mode.

Nevertheless, I would be glad if you introduce this conditional mapping, but I think it is better if you do it yourself, since you know best how it fits into vim-rsi.