It looks like the data needs to be normalized? People might be assuming different thresholds of happiness, the same way the common response to "are you ok?" is yes, and continues to be yes until someone is absolutely miserable.
I agree with you, and guess most people would. These surveys likely have linguistic and cultural nuances that I imagine are extremely hard to normalize for.
Don't measure people across regions via these sort of metrics.