The chi-square and binomial tests compare an observed categorical distribution with a theoretical distribution.

Are the values entered the exact number of objects or events ?

The results can be interpreted only if you entered the actual number of objects or events. The results will be meaningless if you enter normalized values, rates or percentages.

Do the expected values come from theory?

The whole point of this analysis is to compare an observed distribution with a distribution expected by theory. It does not compare two observed distributions.