Problems of endogeneity [1] occur for many different reasons, but are present (to some extent) in nearly every empirical macroeconomic study. They arise whenever there is a circular dependency among the parameters you're trying to measure (economic growth is dependent on education is dependent on economic growth). While we do have statistical tools to help control for endogeneity biases, like instrumental variable regressions [2], we often don't have enough data points (we can only measure economic growth accurately for 100 some countries for 60 some years).
All of this wouldn't be a problem if Macroeconomists could run controlled experiments, but this isn't really possible for fairly obvious reasons (Good luck convincing a country or three that they need to change how they run their country to validate some economic model).
So sure, Macroeconomic models may be somewhat invalid because of assumptions of rational actors, but I'd argue that Macroeconomists have looked the other way, not because there wasn't anything better (as you've said, behavioral economics is making a lot of advances) but because there were larger issues that need to be addressed.
The Microeconomists and game theorists can work out the rational actor problems.
The field was advanced tremendously by people like Daniel Kahneman who won the Nobel prize in economics in 2002. The field of economics is being transformed, but as with all big changes it takes time.
Kahnemans nobel prize lecture is worth watching if you want some insights into how people are irrational in predictable ways: http://www.nobelprize.org/nobel_prizes/economics/laureates/2...