Understanding Society User Support

Dear Understanding Society team,
I am wondering why there is many missing/inapplicable responses to oprlg in later waves? Whilst in Wave 1, no inapplicable answers are coded, in later waves, this is between 80-90%. Even when merging with Wave 1 responses, or using the fed forward variable, I still have around 50% of values missing for adults at different waves. I am working with a created dataset of children and their parents, and using either the mother's or father's religion still yields a lot of missing responses. I will appreciate any guidance on how to figure this out; and how to properly code no religion vs missing values.
Thank you so much,
Michaela