Avalon-ST vs Descriptor/Data in pcie ip 9.1

In the PCI Express Megafunction, two options are provided for the interface between ip core and application:

Descriptor/Data and Avalon-ST, and the latter is recommended by Altera in the User's Guide.

However, as I read through the VHDL code of the chaining dma example using Avalon-ST, I found that, we first interface with ip core using A-ST, and then we use a module to change the A-ST to Descriptor/Data. The datapath is :

core --> core to A-ST --> A-ST to D/D --> App

And if we use D/D instead, the datapath is:

core --> core to D/D --> App

From the app layer's perspective, both options provides the same interface. So why not use D/D directly and bother to use A-ST ? Is there any advantage we can make from the Avalon-ST?