From my understanding, you do not need to do two apps for those components, it is showing you the equivalent if you decide to deploy in an isolated service plan. i.e. You would use I1 for isolation instead of S1.
In general, these are guidances as minimum recommendations and you can change any of them to be more powerful or less powerful based on the needs of your particular situation. You might need less power on your CD, but more power on your authoring, for example.
The most critical piece is ensuring you do performance testing of your application on the desired end state infrastructure to make sure that the choices you made will perform. The benefit of PaaS is that you can be quite flexible in adjusting your infrastructure after initial provisioning to tweak to where you find bottlenecks.