I don't have an etymological research about the origins of that, but I can understand that using the term "Error" may not be precise in all situations; also as almostSharepointMaster mentioned, it is better to think of the error and the exception thrown as separated entities.
When you're in a high-level programing language, it makes sense to assume that an exception is always caused by an error, although I agree also with dasblinkenlight that even then an exception not always is the consequence of an error. I, for instance, use exceptions to terminate threads collaboratively.
The first time I saw the term "exception" was in the 80386 Assembly manual. I remember that when I saw it looked instantly natural to me. To call that an error would not be correct, because there are no errors in Assembly; there are simply conditions the processor can't deal with (if that's an error - from the programmer, the user or the system - well, the processor is completely agnostic to that). I don't know if Intel really originated the term or not, but maybe...
XYError
-- for instance, in Python. – delnan Aug 18 '12 at 23:45