Understanding Society User Support

I am working with the Wave 6 Indresp file using Stata. I have limited the sample to respondents who have a paid job regardless of whether they are employees or self-employed as well as by an adult main cross-sectional weight greater than 0 using the following code:

keep if (f_jbhas 1 | f_jboff 1) & (f_jbsemp 1 | f_jbsemp 2) & f_indinub_xw > 0

From the tabulation of the variable f_jbrise (pay includes annual increments), 3,073 respondents were not eligible for this question and this can be assumed to be due to that they were self-employed and NOT employees as f_jbrise only applied to employees. However the frequency for the self-employed in the sample is 2,584 which leaves 489 respondents who were not eligible unidentified. This is confirmed by a cross-tabulation of f_jbrise and f_jbsemp (employed or self-employed in current job). I have tried to identify these 489 respondents by cross-tabulating possible variables that might include this group without much luck. However a cross-tabulation of f_jbstat (current economic activity) and f_jbsemp for these 489 gave some contradictory results:
1) 468 of the 489 respondents in paid employment were employees but were classified as not eligible for the f_jbrise question. Others (3 retired, 1 maternity leave, and 12 full-time students) were also employees.
2) There is an overlap of mutually exclusive categories (5 self-employed respondents are also employees).

Can you help me understand who these respondents are, why they were classified as not eligible for the f_jbrise question and how they should be treated in an analysis? I have attached a word document with some notes and some tables of the outputs.