The performance difference will be in reading the metadata about the layers and obtaining the data to display. If you are creating a map in code and adding data from a feature service, it takes time for that feature service data to get to your iPhone/iPad. On the device, the data contained in the mobile map package is right there and data access is really fast.
The only thing to consider between reading feature data from a feature service and reading it from a MMPK is that a feature service can generalize polygon and polyline data suitably for display so as not to download more vertices than can be displayed. Polygon and polyline feature data read from a MMPK is read in full and not generalized on the fly.
If a mobile map package works for your app, there's no reason you shouldn't use it if you want to. However, web maps allow for the map content to be updated as needed. You need to work out how to distribute your mobile map packages yourself.
Note that as of 100.9, mobile map packages can include "online layers", which can be used when there is a network connection available, so you can get the best of both worlds if need be. There's of course a local storage overhead of using mobile map packages.
Does that answer your question?
