I'd say leave the source files as Win-1252 so external tools can work with it.
...Though nowadays Unicode would be preferred, while some 32 characters in the middle don't match between Unicode and Win-1252, because the first 256 characters in Unicode match ISO-8859-1 instead. Oops.
(I found it was easier to write external tools of my own to handle SCI 1.1+ UTF-8 support than to convert SCI Companion. Go fig.)