I have a 120 item, 18 facet, five factor model I am attempting to validate. I have attacked this by attempting to develop single congeneric models for each of the 18 dimensions, but am at a loss as to where to go from there using SEM. Can I impute facet scores and work with the facets or should I continue with the items within the model (which is very messy)?
Your model is way too big to ever get a reasonable answer from SEM if you use individual scores. You can calculate (not sure if you mean that by impute, but imputation is usually related to missing data). Most (every?) program will let you save a factor score, or you can calculate them, and then feed these into your confirmatory factor analysis.
What's the sample size?