I don’t understand the topic or study, per se. But in general, not liking the results is not a good reason to adjust data, or abort properly conducted work. Sometimes a failure or problem in results can indicate issues not expected. Perhaps there is another reason why results were unexpected and poor. If one skews results to please reviewers, that may be considered unethical for a researcher. Results are supposed to be unbiased. Perhaps better to try to explain the variation or poor result, consider retesting or replicating if needed.

Hey William, thanks for your answer. Do you know about latent class analysis? I'm not trying to skew my results to please anyone. In contrast, I would prefer the class solution with the low average latent class probability in one class and I could explain why two classes might not be exclusively distinct. However, I wanted to know if this approach is okay?

Also, I did various sensitivity analyses in order to check my model for stability, but the average latent class probabilities were low in all analyes and they've been also low in a 2-class- or 4-class-model.

No, as indicated, I am not versed in this type analysis. Multivariate Analysis is probably a relative, but this approach was not mentioned in my statistics courses that I remember. It seems to involve assumptions, design, interpretations, etc. to pull out possible behavioral trends or meaning from perhaps logics, carefully designed questions to participants. I think this type approach would sometimes result in surprises that were not suggested, and a degree of uncertainty that may make the statistical designer or analyst uncertain or question results. I can see that slightly different questions or classes could affect result or interpretation. I am not sure if the “findings” could be statistical level certainty, or something like trends. I am reminded of a TV series I like called Bull, where he tries to read people behaviors relative to court juries. I agree there is probably some untapped information to be gleaned by those with the training, expertise and insight to pull it out.