UTF7 and UTF8 conversion by MultiByteToWideChar and WideCharToMultiByte is strictly Unicode to Unicode and none of the flags are honored except for WC_ERR_INVALID_CHARS.

I've noticed a few cases where people expect MB_PRECOMPOSED behavior because the documentation says "This is the default translation option", however for UTF7 & UTF8 the Unicode is just converted to a different form, it isn't translated.