Does anybody have a clear way of how to separate these from each other?
Recallability Trap - giving too much probability weight to recent (often negative) extreme events mostly in forecasting. Eg. given last global finance crisis one may conclude there is a significant probability that equity market will not recover for many years.
Availability - proceeding upon first available information without considering other sources of information. Eg. you read in newspaper that energy stock will outperform given some forecasting and you have decided to invest upon this information only.
Representativness - partially concluding that one presumption or even known characteristic of some thing determine your decision. Eg. Good companies means good stocks and are always good opportunities to invest.
Recency - some recent event (might not be extreme), positive or negative has greatest probability weight in your decision making.
Ok, so the first is only related to forecasting while the other are biases?
Yep, according to CFA curriculum. In real life maybe not.