There is little point in localizing just a part of a currency presentation and using non-localized or wrongly localization notations otherwise. It may confuse, and it gives the impression of half-hearted localization efforts. The CLDR database contains information about the placement of currency denotations, too. They would not have included it if they thought such things can be left unlocalized.
If you can successfully localize some user interface or some presentation of some data, the odds are that currency symbols used in a locale are available. This has changed to some extent in recent years, since symbols like the new character for Turkish lira and the Indian rupee symbol are new and generally not supported in fonts, and they may cause other problems as well.
So it may be better, in some cases at least, use currency abbreviations or names, especially since symbols like “$”, “£”, and “¥” do not unambiguously indicate a single currency. The CLDR data contains localized abbreviations and names for currencies.
The three-letter codes are meant for international banking business and for data interchange between computers, not for normal human consumption. In a user interface for common people, they should not be used, except perhaps as the last resort when everything else fails. But there’s not much reason why things should fail. If used, these codes follow the normal language-dependent placement rules (e.g., 10,00 EUR in French).