Is your problem with the implementation of solutions, or recognizing which solution to apply to a problem? Interviews are testing for both, but really the most important aspect is recognizing the stated problem as an existing one. That is, you must be able to translate a convoluted description into a well-known algorithm, like "Oh, this is asking for min-cut/max-flow!" Formally, this is called reduction  -- reframing an unsolved problem into one you know how to solve. Not only is it an important skill in day-to-day work, but the better you are at it, the more efficient you will be in interviews, where time is precious. You don't want to spend 10 minutes deciding how to approach the solution, only to realize 20 minutes into writing it that it's the wrong approach.
It sounds like you have a really solid base of knowledge from all that practice. Perhaps you could work on leveraging that knowledge more by improving your ability to quickly pull solutions from it.