Hello. I have a continuous outcome variable, binary endogenous variable, and continuous instrument.

I have estimated the effect of the endogenous variable on the outcome variable using etregress with maximum likelihood and with a control function. I've also estimated it with 2sls. The control function and 2sls
results are very similar, but etregress with maximum likelihood is very different. Why is that occuring? From reading the documentation, my guess it that the maximum likelihood approach requires parametrizing the unobservables as
jointly normal, while the control function only requires univariate normals? Is that right?