I was fighting the same problem today.
The regex in eregi_replace() doesn't seem to be doing the right thing at all.
I've added this line to google_base.php just after the call to eregi_replace() in the sanitizer code, and it seems to be cleaning things up rather nicely.

Code:
$str = iconv("UTF-8","UTF-8//IGNORE",$str);
Cheers,

/Al