Understanding Society User Support

Dear Sir/Madam,

We are currently look at some of the self-completed measures in the UKHLS Wave 3 data; our sample is restricted to those who accepted and participated in the self-complete and it excludes proxy respondents from the indrep file.
Following the user guide, we have applied the lowest level (1) self-completion weight. However, the introduction of this weight substantially (and quite counterfactually) changed the descriptive statistics on some key demographic characteristics.
We further check into the x-sectional self completion weight and found that there are more than 20% of the sample weighed '0'.
(We also tried with using x-sectional individual respondent weight (indinub_xw), and the results from which seems more 'normal'.

Could you please advise us as to what should we do with the weight? Which weight is appropriate? And would it be appropriate if we use indinub_xw (individual weight instead of self-completion weight) with self completed measures included in our analysis?

Thank you very much in advance for your advise.
Very best wishes,
Yang (Hu)