If respondents are answering randomly, their RLH scores (shown for each respondent row of HB scores) from HB analysis will be relative low. For example, if you showed 4 items per screen the null (by chance) RLH would be 1/4=0.25. And, since HB tries to fit the data (even if random), the RLH from random data for 4 item shown per screen could be somewhere around 0.25 to 0.35.
