chi-square results not equal if expected proportions used instead of counts in R

Question

testing goodness of fit between two samples from the same fact in different months. I want to know if results in 4 categories from September are similar/different from Octob