st: estimation of heteroskedasticity 3SLS using cmp (conditional mixed process) estimator

[Umid Aliev <U.Aliev@leeds.ac.uk>]

Dear Stata users,
I am trying to estimate a simultaneous equations system using H3SLS. As it was advised on previous correspondence in Statalist: www.stata.com/statalist/archive/2009-09/msg00426.html, I am trying to use cmp command by D. Roodman (2008). The syntax used is as follows:
cmp (y1=y2 x2 controls) (y2=x2 x3 controls) (x3=z1 z2 x2 controls), ind($cmp_cont $cmp_cont $cmp_cont)
The controls are set of dummy variables, identical in all equations. x3 is not a separate equation, it is just instrumented as it is endogenous in y2. However, the results seem to be wrong, as they are identical if I run the command as given above or only for one equation, like
cmp (y1=y2 x2 controls), ind($cmp_cont )
In my inderstanding, the estimates should differ since in the first case I use (I think) information from the whole system, while in the second case only one equation.
Moreover, inclusion of robust option does not produce heteroskedasticity-robust st. errors, although it does not indicate any syntax error. It just ignores it. Standard errors are exactly the same as without robust option. 
Any help on these matters will be greatly appreciated.
Many thanks and sorry for a long post,
Umid  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/