Is your objection to the number in the sample or to the use of registered rather than likely voters?
Both actually, I am the type that only considers a poll valid if it includes most of the people who it affects, I.E. all the voting public in this case.
However, for those that believe in political polls, a much bigger sample, 900 is just assinine for this type of question, and "Likely voters" instead of registered voters.
Quite a few registered voters never vote.