Robust indices WLSMV availability sem.mi()

[Laura Wade]

Hi all,

I am fitting a series of models to multiple imputed data (65 imputations, sample = 4,982) using lavaan (ver 0.6-21) with WLSMV estimation and ordinal indicators. I have first run a measurement model fitted with cfa.mi(), followed by a full sequential mediation SEM fitted with sem.mi(). Pooling method is D2. 

I have confirmed the following:
- Both the measurement and structural models converge on all 65 imputations
- The polychoric correlation matrix is positive definite for the full SEM (all eigenvalues > 0, confirmed by lavInspect)
- No pooled Heywood cases

My issue is that cfi.robust, tli.robust and rmsea.robust are available for the measurement model via cfa.mi(), but are NA for the structural model via sem.mi(). I'm using fitMeasures to get the pooled fit statistics. Looking at this lavaan manual https://yrosseel.r-universe.dev/lavaan/doc/manual.html it seems that robust indices should be reported when available, however as i want to compare my measurement to my structural models (and be consistent in what im reporting),  I'm not sure which indices I should be using.

My specific questions are:

1. Is it normal for cfa.mi() to provide robust indices, and for sem.mi() to not? 

2.  Is there anything I can do to obtain robust indices for the full SEM? I've tried checking robust indices specifically, but get NA in all scenarios. 
3. If there are no ways to get robust indices with sem.mi(), given that robust indices are available for my measurement but not for the structural model, should I report standard or scaled indices for both models for consistency? Or is there a preferred approach? 

Thank you in advance! I'm new to SEM/Lavaan so it's all a learning curve