[Chapter 7 - Exercise 1] Ensemble Learning and Random Forests

The question says if we combine 5 different models trained on the exact same training data and they all achieve 95% precision. Is there any chance to combine and we can get better results?

Shouldn't the evaluation metric to be considered be `f1 score`, wouldn't precision be a loose metric? There would be a possibility of high precision and low recall - The classifier would only return few results which would be correctly classified as positive. However it would miss on many actually positive results.