I have had this experience and I gather it is not unusual. First of all, take a look at the factors that are loading on each subscale. Is there an underlying clustering that points to a characteristic. You can drop some of the lower connecting factors to make each subscale more focused. You can add more data and repeat the factor analysis to see if that clarifies the loading. Also, while looking through what factors are loading, consider if inferential intuition and affective intuition are too similar to warrant two separate scales.

Here's a link to a discussion about this issue that might be helpful. https://stats.stackexchange.com/questions/58757/what-to-do-with-a-variable-that-loads-equally-on-two-factors-in-a-factor-analysi?noredirect=1&lq=1

Translation is a tricky thing, because you're dealing not just with different words for the same concepts but also with words that refer to subtly different concepts and the underlying cultural differences that intersect with language. So you should not be surprised to find that some aspects of a scale's structure change, especially if the linguistic/cultural groups are quite different from each other.

That said, it's worth checking the following. Did the original authors use the same statistical procedures as you did? As you know, different rotational procedures can lead to surprisingly large changes in the results. And what were the factor loading in the original study? It could be that the items in question were always a bit "muddy," with strong secondary loadings on the factor where you obtained the primary loadings.

Beyond that, you'll have to wait for someone with real statistical expertise to weigh in!

Thank you for your quick responses. I will take all of this into consideration in deciding what to do with those items! Again, thank you a lot for your help. I am really really grateful!!