Logistic regression


in my study I've conducted a logistic binary regression (personality trait - 5 scale variables, dependent variable is nominal binary). When using enter method, there were no significant predictors (omnibus and Hosmer Lemeshow both not significant). But using stepwise LR method I got a fitting model (omnibus significant, Hosmer Lemeshow not significant), consisting of only one predictor and with Nagelkerke R value being only 0,03... I'm not quite sure how to report this - that there is a fitting model that explains almost no variability?

Another dilemma I have is that the two groups are significantly different in size (530 vs 26 samples), is it even acceptable to use regression in this case? How does this size difference affect the results of analysis?

Thanks in advance.