The wide-spread differences between RNA and DNA sequencing are most probably artifacts and not RNA editing events as suggested previously in the Science paper, except A to I editing (an I is read as a G) and C to G
The reason why they have such a large number of artifacts is most probably the presence of systematic mapping errors. Reads coming from RNA-Seq will rather be mapped by a mapping software in one piece to the wrong location (even if they map with few mismatches to the wrong location) than be spliced between exons. Thus lot of reads will map to non-functional pseudo-genes, for example, or some other homologous locations with mismatches. The sequence will be different from the underlying DNA sequence and thus it will be called a RNA editing event.
Also the confirmation with PCR may be erroneous as the primers may be designed poorly and may amplify 'the wrong' homologous sequence as well. This may give the false conclusion that the DNA sequence is different from the RNA sequence at the gene locus.
More details can be found here: