Binary logistic regression (SPSS) - input & output

Treatments A and B have been compared in a classical RCT. The primary outcome is dichotomous - remission. Treatment B, the experimental, resulted in significantly lower remission rates.

In a binary logistic regression analysis I used treatment and three additional factors as covariables (or whatever nomenclature SPSS uses). Treatment and covariables were added as "singular factors" and as interactions:
Covariable X
Covariable Y
Covariable Z
Treatment * Covariable X
Treatment * Covariable Y
Treatment * Covariable Z
(all as categorical variables)

The regression was done in two steps. The second modell could significantly better predict the outcome (with a Nagelkerke R square of 0,22).
One of the interactions was significant.

As I have a modest understanding of the method, I have some questions:

Have I, regarding the input, asked SPSS to calculate the regression the correct way?

In the final output, treatment is no longer significant - the p-value is actually close to 1. Is this just an "anomaly" of the method? The significant interaction between treatment and one of the covariables (being old or being young) is significant as remission rates are higher in one treatment arm if old, and very much higher in the other treatment arm if young.