Found a good article which explains the differences between the usage of the EL expressions like
${expr} (immediate) and
#{expr} (deferred)

Here is the link

Here are some of the important points from the article (in case if you don't want to follow the link..

The primary limitation of the JSP EL (expressiongs like ${expr}) is that its expressions are evaluated immediately, meaning that the JSP container immediately parses and resolves the expression when it processes the page and returns a response.

Immediate evaluation is sufficient when a JavaServer Faces page is rendered for the first time. However, after the user enters values into the UI components and submits the page again, those values are converted, validated, and propagated to server-side data objects, and component events are processed. In order to convert data to the proper type, validate it, store it in an external object, and process events in an orderly fashion, the JavaServer Faces life cycle is split into separate phases to handle each of these tasks. Therefore, it must be possible to evaluate expressions at different phases of the life cycle rather than immediately, as is done in JSP technology.

Another problem with the immediate evaluation of expressions is that it is done in read-only mode. JavaServer Faces components need to be able to get data from server-side objects during the rendering phase and set the data in the server-side objects during postback.

Finally, JavaServer Faces components need a way to invoke methods on server-side objects during various stages of the life cycle in order to validate data and handle component events. JSP functions are not sufficient because they can only be used to call static methods defined in a TLD file; they cannot be used to dynamically invoke public methods on objects.