The three characters are UTF-8 BOM mark:
I suppose the file has the BOM included on its own. When you open the file in WinSCP internal editor, it autodetects the UTF-8 encoding by its presence (and does not show the mark as it's not part of a contents). If you force WinSCP to interpret the file in ANSI encoding, it displays these characters as if they were part of the contents. But this has not changed between 5.1 and 5.5.1. Can you please test this once again using 5.1?
If the above does not explain what you experience, I have further questions:
What encoding do you have configured as default in Preferences?
Am I right assuming you are using WinSCP internal editor, right?
What protocol? FTP or SFTP?
Is the BOM present in the file before upload? (the file is 3 bytes longer than its contents if BOM is included).