Have you tried fitting the original expected 5-factor CFA model to your data? What does a 5-factor CFA solution look like? Does it fit and do the item loadings look sizeable? An EFA approach can sometimes be misleading depending on the criterion used for selecting the number of factors and other choices.

If the number of factors really doesn't match with previous research regarding this scale, the reasons may be cultural or related to the translation.

I am trying to cite cultural and translation differences. Item loadings are sizeable. EFA made me reduce from 14 to 12 items.

Does the sample category impact factor structure? Originally, the scale was administered on general population and I administered it on clinical population.

Christian Geiser

It could be that your clinical sample has a more limited score range (smaller item variances, less inter-individual differences) which may result in smaller average inter-item correlations. This could explain the differences in factor structures. I would study the inter-item correlations and compare them to the ones in the original study.

Two points. First, the notion that there are only two or five constructs that influence how a sample of people will answer a set of questions will only be true in computer simulations. Many things influence how people answer questions: outside of simulations ``there is no such thing as `the true number of factors to extract,' since the only possible assumption is that both the number of substantive and the number of error common factors each exceed $n$, the number of variables'' \citep[p.~273]{Cattell1966}. Second, because finding "the true number of factors" is not the goal, instead you are trying to simplify the analytic story with a simpler causal model of how the data arose, ignoring what you think are unimportant (though true) factors for the story as it relates to YOUR sample/problem/context. So, the second point is that you finding two important factors may be appropriate for your sample/problem/context, but of course that depends on the story you can tell with them so requires more than just looking at a few summary fit statistics.

A change in factor numbers can be due to a variety of issues. First it may be that the first confirmatory factor analysis did not reach a true construct validity. Second in the repeated analysis the sample size may have been more or less representative than the first. Third there may have been redundant items in the original dimensions. Fourth, it may be related to the items in the dimensions if they were not truly measuring that concept.

it is important to review where the items loaded and which ones from the original factors are now loading. Sometimes you end up with a clearer set of fewer item in a further analysis because the additional previous dimensions are now not distinct but are integrated together into a single one. Try a further Re analysis with a two factor forcing to see if this was present in the first analysis.