The part about InputStream versus File for performance is probably bunk, I agree. But if you pass either of those two things to an XML parser, it will determine the document's encoding from the document itself in the way specified in the XML Recommendation. There's no need to make your own Reader to specify the encoding; in fact this will backfire if you specify the wrong encoding.
The problem is this: the actual error message says
Character reference "" is an invalid XML character.
This is true, and it has nothing to do with Arabic because that character is nowhere near the Arabic ranges of Unicode. It's the ASCII "form feed" character and it just isn't allowed in an XML document. It's better not to try to delete those characters before you parse the document; the correct approach would be to not put them into the document in the first place. So contact whoever produced the document and explain the problem.
The XML Recommendation is
here and you might want to read the relevant sections (chapter 2.2 for example) so you understand them before you try telling somebody else their XML document is malformed (which it is).