I think the characters thing could benefit from a dropdown setting with the following options (possibly excepting the second):
* Generate character variables, and include those character variable declarations and uses in the output. (This would make a file that runs, and the user would be able to adjust the character variable assignments if they cared to.)
* Output character names in quotes so that they'd be strings rather than variable names. (This would make a file that runs, but with no formatting on speaker names; on the other hand, there'd be no conflict with already defined character names.)
* Use variables for character names, but do not generate a definition for them. (This would make a file that does /not/ run unless the user has generated a separate file with character variable declarations or adds the same to a generated file after the fact.)
The first two options would be useful for someone who wanted the generated files to "just work", and the third would be useful for someone who wanted to run the conversion tool iteratively, but didn't want to repeatedly respecify the formatting for character text output.
I'm inclined to think that one of the first two options should be the default setting. A new user who uses the default settings on the sample Twine file should be able to end up with something that works as expected in Ren'Py (otherwise, the new user might think the tool is "broken", even if the tool is performing exactly the intended behavior).