I am trying to find out the effect of the interaction terms in my model. but i keep getting conflicting p values when i use # compared to when i use ## in stata.

both variables are categorical, my response variable is also categorical.

when i use the # in stata, below

Code:
melogit pass age#class || group:
my output shows only the interaction and the interaction is significant

when i use ##,below

Code:
melogit  pass age##class || group:
my output shows the main effects and the interaction and the interaction terms are not significant.

thank you for your response.