What you have to be wary of is the “affirming the consequent” fallacy.
If A, then B
There is DNA that is transcribed into functional non-coding RNA. However, just because DNA is transcribed into RNA does not mean that the DNA codes for a functional non-coding RNA molecule. To use a bad analogy, there are air filters that pull in air and bind dust particles on their filter. If I smash a TV into a million little pieces and pile them on the floor, those pieces will bind a few dust particles as they float by. Therefore, since the busted up TV binds dust particles just like the functional air filter, then the busted up TV must be functional as well. This is the (bad) logic that ENCODE used to justify RNA transcription as an indicator of function. For some “functional” RNA molecules in the ENCODE study they were found at less than 1 copy per cell, and yet they still called them functional. To make things worse, ENCODE also used several cancer cells lines that are known for leaky RNA transcriptase activity.
As Larry Moran over at the Sandwalk blog puts it, non-function is the null hypothesis. You need evidence for function before you can say that an RNA molecule or even a translated protein has function. One of the most well recognized pieces of evidence is sequence conservation, and only a small fraction (~10% of the human genome) shows evidence of conservation. Can DNA still have function without showing evidence of sequence conservation? Yes, that is possible, and I wouldn’t be shocked if 1% or even 5% of human genome was functional but showed no signs of conserved sequence (although 5% is really pushing it). Is it possible that 70% of the human genome that shows no evidence of sequence conservation somehow still has function? I think that is fantasy, and yet that is what ENCODE suggested when they said 80% of the human genome has function.