I have come across two different definitions for a 'irreducible set of states' of a Markov chain. 

Definition 1: A subset of states $A$ of a Markov chain is irreducible if it is possible to access (in possibly more than one step) each state from the other.

Definition 2: A nonempty set of states A is closed if x∈A and x→y implies y∈A. A closed set A is irreducible if A has no proper closed subset.

My question is the following:

(Q):  If A is a set of transient states that are accessible from each other, does it also mean A is an irreducible set of states?  


Why do i care?
--------------
The following is perturbation bounds paper for finite Markov chains.
http://www.jstor.org/stable/3212261

The paper claims the perturbation bounds apply to two Markov chains if both of them have a 'single irreducible set of states which don't necessarily overlap'. Referring to (Q), if a Markov chain has a transient set of states that communicate with each other  and another closed set of states that communicate with each other, does the Markov chain have a 'single' irreducible set of states or 'two' irreducible set of states?

Example: 
-------

$P = 
\begin{bmatrix} 0 & 1/2 & 0 & 1/6 & 1/6 & 1/6 \\
 0 & 0 & 1/2 & 1/6 & 1/6  &1/6 \\
1/2 & 0 & 0 & 1/6  & 1/6 & 1/6 \\
0 & 0 & 0 & 0 & 1 & 0 \\
 0 & 0 & 0 & 0 & 0 & 1 \\
 0 & 0 & 0 & 1 & 0 & 0 \\
\end{bmatrix}$






Does this Markov chain have a single irreducible set of states {4,5,6} or two irreducible sets {1,2,3} and {4,5,6} ?