It is true that this warning applies to all macro, not just undergrad. But by the time you get to the grad level and beyond you already understand (or should understand) the limitations of the models and the difference between academia and real world. At undergrad level you're still likely to be seduced by the simple narratives that these models offer.

Are complex macro models how the world actually works?

Nope! All models are huge simplifications.

What are our most successful macro models, and how successful have they been?

A controversial question!

The conventional approach in (academic) macro is to build (relatively) simple models that can match particular stylized facts. Thus we have lots of models that can predict certain patterns in the data, but don't pretend to explain everything. Some people think we shouldn't do anything beyond this! (See Caballero, Pretense of Knowledge Syndrome)

Other people do try to build models that can match all the data. A standard cite is Christiano, Eichbenbaum and Evans (2005), and for another approach see Smets and Wouters (2007) (they even call themselves Bayesians!).

Then there are macro forecasters who try to accurately predict the future using non-theoretical statistical models. An example of this would be the work of Frank Diebold at Penn. These models can do a lot better than the above at predicting future data absent changes in the policy regime, but are presumably less effective at predicting the effects of novel policies (see the Lucas Critique).

My own opinion is that if you want to predict the next data point, use a forecasting model, but if you want to know the effects of a new policy, your best bet is to rely on simple models plus judgement. Good economists know more than any model!