The survey statistics have a major bug that leads to wrong graphs/values. Let me explain: We've had a questionnaire with 178 participants, but of course not every person answers every question. So the number of participants varies from question to question. If you look at the attached screenshot and count the participants you'll see that 150 persons answered the question, but the percentage values assume the whole number of participants (178). That leads to wrong graphs and wrong percentage values. For clarification I added the correct percentage values.
The fix is trivial: Use the actual number of participants instead of assuming the overall number of participants is stable.
Note: I didn't test it in OO 10, but the error is probably there, too.
See comments: we agree on fixing it by adding a bar for the user count that did not answer the question. In a perfect world there would be a toggle to switch between the two views, but for now we go with the additional bar as default.