It is insufficient because it supports only selection between alternative resources. Adaptation is more than this.
It does not necessarily have the appropriate granularity. There is no longer a guarantee of a one-to-one relationship between a page an author creates and the page as experienced by a user. One authored 'page' may turn into many perceivable units on a small device.
It does support situations where adaptation is just selection. Hence the image example.
An adaptation processor is building a delivery unit. It needs to provide a reference to an appropriate version of an image for use by the requesting device. It selects an appropriate representation based on the metadata available. This might include width, height, color depth, encoding etc. In addition, an author might express intent about the destination for these versions, again in terms of metadata. Note that all of these characteristics, including the author's intent, apply to the entire resource. Adaptation is selection in this case.