I don't have a better answer, but just beware.  The certification exam
is well-known for its very detailed questions about language features,
requiring very precise consideration of the answers.  The SCJP exam does
this very well, regardless of whether you think this kind of language
knowledge is useful or not.  However, when people who are not experts in
the Java language try to replicate this question type, the results are
atrocious.  We have frequently seen people in the past post SCJP mock
exam questions for which the only possible correct answer would be "I
don't know" or "None of the above" or "A, B, and sort of D".  This kind
of situation doesn't prepare you at all for the SCJP exam.

I'd advise that, at least once, you actually read the language
specification before taking the exam.  Even if you don't understand
everything (and if you DO understand everything, you're doing better
than 98% of people... or 99.98% for the third edition) (*), it will
still clear up some of the fuzzy impressions you get from bad test prep
sites.

As for the value of the test itself, I actually am rather fond of it at
least in the form it took when I took the test quite a few years ago (ca
2001) -- not because it asks the kind of thing that should occupy a good
Java programmer's thought processes during a day of programming, but
because many of the questions that it asks assume the kind of background
that is necessary to work comfortably in Java without stopping to think
about how some operator or parameter passing works.  Furthermore, it
does ask about things that any good programmer should be capable of
understanding... and the only way a test can measure that is by telling
you to try to understand it, and then having you show up and take a test
to demonstrate that you succeeded.  Even if you forget a lot of the
minutae of the language that you've memorized for the test, you have
still demonstrated that you were capable of correctly understanding
them, and that you have the basis knowledge that's required to see the
significance of the question in the first place.

(*) And 80% of statistics are made up.  Of course those aren't real
numbers; just making a point.