Towards that channel argument:
1.) Hans Brinker is right. If you pick two channels from one sensor and one channel of another sensor for voting you are distorting your probability of occurence of certain fault modes to a degree where you cannot make any useful statments of the failure rate of the entire system.
2.)
AFAIK the two channels are processed by one air data unit and are used for diagnosing electrical faults. That's what they are good and useful for. Hooking them up to different FCUs will either deprive you of that integrity check or things get an awful lot complicated, see 1)