What do you mean by "is this normal"?

If you refer to such observations in other (similar!!) data sets you will need to know other (similar!!) data sets. How many do you know? How similar are these to yours?

Another possible approch to find out if your finding is "normal" is to generate (many) similar data sets under resonable assumptions of the data-gererating process and to have a look how often such findings do occur. The problem, for sure, is to pin down a reasonable process, since this might be just what you want to find out from your data. That's like chasing one own's tail...

Fabio,

if you use t distribution, you need degree freedom. I think here you used z, standard normal. For z=1.74, two tail test p-value is about 0.08.  For interaction with such p value, you do need it in the model in general.

You didn't give  factors for interaction, If one is categorical, say G, one is continuous say X, it means that the regression coefficient of X changes by G. Say G is gender, it means the x effects are different between males and females. the difference is 12. if G is continuous, then it means, when G increases 1 unit, the X effect increases 12.

If this (12 or 7)  is high or not in magnitude, it depends on the measurement of dependents variable. We can't make judgments without knowing the data distribution.

Many thanks for your answers. They gave me some new insights about the topic. I am going to think deeper on it.

Consider the following scenarios:

A and B are both insignificant

A is significant, B is insignificant; or the other way around

A and B are both significant