The file doesn't contain <93> or \x{93}; it contains a byte with a value of 0x93. This is garbage (where found) when UTF-8 is expected. The two programs simply handle this garbage differently.

What you have is probably a file encoded using cp1252 rather than UTF-8. «“» and «”» are encoded as 93 and 94 respectively when using cp1252.

Convert the file's encoding from cp1252 to UTF-8 before using it with tools expecting UTF-8.

iconv -f cp1252 -t UTF-8 file.cp1252 >file.utf8
[download]