I recently found my old cube a few months ago, and had a hard time remembering the algorithms that I used to solve it thirty-odd years ago.
Then I came across Petrus' web site (http://www.lar5.com/cube/index.html) and was able to solve it again. An ingenious approach.
After getting to know the method, I discovered that I often could save some turns in step 4 if could put in a different 1x2x3-block than one of those two you are supposed to build at this step. Although that would mean that I had to solve the final layer with one of the sides turned, and I would (usually) get an extra turn at the end (sometimes, that turn would actually coincide with or even be the opposite of the last side turned of the last algorithm - thus no extra turn, or even a turn less).
After practicing this a number of times, it turned out to be not too difficult.
Then I tried to do this from step 2 as well, doing step 4 as normal. By carefully studying the edges doing step 3, that didn't turn out too complicated either.
Then finally, I tried to leave a face turned both in step 2 and step 4, leaving two extra interleaving turns at the end. I have to confess that I have failed several times at the final layer doing this, but I have also succeeded a number of times (recall, I'm just a hobby-cuber, a pro should be able to do this easilly).
With the abundance of cube-solving sites in the Internet, I'm puzzled why I can't find any description of this "twist", it's so simple.
The method is maybe too complicated for speed-solvers, but for those that are into solving it in fewest possible turns, it is definitely a turn-saving technique.
Are there any of you out there using this method, and what is your experience with it?