Problem Fitting SEM model with Reflective Latent Construct

I am trying to fit a MASEM model where:

- "index" is a reflective latent construct with five indicators (WE, IE, FK, EE, & NVI)
- "OB3" is a variable that is predicted by index
- "OB1" is a variable that is predicted by OB3, such that the effect of "index" on OB1 is fully mediated by OB3

The graphical representation of the model I'm trying to run is in the attached Rplot

The issue is that when I estimate it using the attached code, the plot of the Actual Model used in the empirical analysis doesn't look like the one I'm trying to run. As you case see in the "ActualModel.png", both OB3 and the indicators of "index" (WE, IE, FK, EE, & NVI) are below the latent construct "index", while in the model I'm trying to estimate, OB3 goes below the "index" latent variable, and the indicators of "index" go above the "index" variable.

I've attached the code and the disguised data (simulated data results in a problematic hessian matrix). I would really appreciate any guidance you can provide regarding how to restructure the RAM model to be able to run the desired model. Thanks for providing such great software for conducting meta-analysis!

Thanks,

Sergio