I am working with a data set which I did not collect, but a (recently death) college. The idea was to examine validity and confidence of a questionnaire. However, the sample he got is only N = 122. There are 13 items. Reading some literature, I found that, although some literature says the questionnaire reflects a one dimension variable, other studies concluded that there are three dimensions (factors), and other studies find even other structures. I conducted EFA, and although I found three factors (using maximum likelihood and oblimin rotation, and also using PCA and varimax rotation), the items do not load in the same way the researchers who created the questionnaire say. I mean, items load higher in factors which are not supposed. With this perspective I decided to try several things: 1) to conduct a CFA using the pattern coming from the Maximum Likelihood EFA. 2) to conduct a CFA with a one factor model. 3) conduct a CFA with a three factors model. I basically always get the same: CFIs and TLIs > .90, RMSEAs < .05, with 90 CI not including zero, X2/df < 3 (this means, acceptable), but significant X2s (chi squares). I have read that when you have big samples you can ignore this. However, I have a medium to small sample (N = 122). Where could be the problem? I have seen several not so nice results with this scale in literature, (including low factorial loadings, diverse factorial structure) and interestingly in manuscripts they do not always report x2, CIs for RMSEA, etc. I think my results are still worthy to be reported. Again, any idea about why I got significant X2s?