Hi Andrew
The Network Rail feeds are railway-oriented and have a lot of data at a very low level. They require a lot of processing and knowledge of the railway, and are generally accurate some time after the event, but not necessarily before. For example, cancellations can be reported in advance on TRUST (pre-cancellations), but a cancellation message won't appear until the train reaches the location it's being cancelled at.
The Darwin (National Rail Enquiries) feeds are passenger information-oriented and have just passenger services, but also contain forecasts for arrival times, and better information on cancellations, part-cancellations, diversions etc. than the Network Rail feeds. The feed is also the same one that most, if not all, CIS at stations use to provide information.
To summarise - if your concern is accurate and consistent information for passengers, use the Darwin feeds. If your concern is a deeper understanding of movements, or you want to include non-passenger trains, use the Network Rail feeds.
Incidentally, I don't think there's any value in using both for customer information purposes - you'll not be consistent with Darwin and confuse your users when they see an 'alternative' version of the truth.
Peter