My project involv collecting data on experiences that arose in participants' work lives over a three week period. Whenever a certain type of situation occurred, the participant would provide data on a number of variables.Our outcome variable is "identity management strategies" (IMS). We adapted items from a number of existing measures to assess these strategies.

I initially had assumed that the within-person and between-person factor structures would be the same on this set of items. However, this was not a good assumption to make, as it turns out. A multilevel EFA strongly suggested a 2-factor structure at the within-person level and a 1-factor structure at the between-person level. Inspection of factor loadings suggested that we should use the same set of items at both levels of analysis. Thus, at the within-person level, there was a distinction between IMS that did not appear to exist at the between-person level.

Given this finding, I am thinking that a better approach would be multilevel path analysis. At the within-person level, there would be two
outcomes. At the between-person level, there would be one outcome. Does this sound reasonable? If yes, then can anyone suggest a published example of this type of analysis (where the factor structure of the outcome variable differs at different levels of analysis)?

I do not think Multilevel SEM would be appropriate. First, the person level sample size is only 51. Also, one of the two Level 1 subscales is scored with only two items.