I've had sites where chars beyond the us-ascii were boxes or ?s. In those cases, I was on Linux and so was the backender, yet I was the only one to not see the characters.
In order to see them, I needed to tell my browser Western (Latin-1, 8859-1) instead of Auto-detect.
But I see it as a bad sign that someone would have to change from Auto-detect.
Your meta tag is UTF-8 and your server is sending out UTF-8... only thing left to check is, how is that HTML (or the text within it) being saved? If it's sent to you in some other charset and then you're stuffing it into the HTML and re-saving it as UTF-8, there could be issues.
For now, though, have your client check his charsettings in his browser. Likely he can get the chars to render by changing something there, since everyone else is able to view it (me too).
Can this be related to the fact that I am using HTML5 in this project?