I have generated the nine datasets according to this document for AES-128 and testing these for randomness using NIST STS. However, the results are inconsistent. Sometimes there appear a * in the result files and sometimes it doesnot. Even the result files show different values on multiple testing of the same file.
Why the behaviur is inconsistent? Most of them times, test fails with a mrginal value e.g if passing crietria is 120/128, then result is 119/128 with a *. Can this marginal failing be ignored?