I think what they are trying to say is that if you take the 1st image and then split the layer-2 access switch into several access switches this affects the behaviour of the topology. Introducing additional access-layer switches switches means the STP topology is more complex (I admit probably not much). It also introduces unicast flooding which you would have had already but you would never have noticed it as the only device performing the flooding would have been one of the distribution switches (typically the HSRP standby switch) and the flooding would have only occured on the downlink. If a second layer-2 switch was in the mix you would have seen flooding here (this isn't an issue if you are using stacked access layer switches such as 3750 or 2960S).
I suggest you search for unicast flooding to understand this as it can be a big issue in networks with VLANs stretched between access layer switches.
However, after just writing all that I now think what the notes are saying is that with the routed access-layer design you cannot have common VLANs between access-layer switches as your uplinks are no longer VLAN trunk ports carrying multiple VLANs but are now routed ports (/30 or /31) and using a routing protocol. In reality you can still stretch VLANs between your access-layer switches by making the uplinks trunks and having your P2P routed VLAN and then a common VLAN you want to stretch. This is something you SHOULD NOT DO though as it destroys what you have achieved by deploying a routed access layer design - you may as well have kept the L2 access layer design.
In my experience there is just no need to stretch VLANs between access switches unless you are talking Server clustering or Virtualisation. User access layers should be routed in my opinion - its IP (v4 or v6) its just as easy to route and much more manageable.
Andy
