I'd argue that pretty much most tornadoes are likely underrated unless there is a Dopper on Wheels scanning them. Most tornadoes go through just rural areas attaining EF-0 or EF-U ratings, yet if they were to hit more populated areas, they would probably come up with much more significant ratings given the lack of DI's over open country. A good example would be the Morton, TX tornado this year, it likely was stronger than EF-2, however due to its slow motion (nearly stationary) and very rural area, that was the highest DI available to rate it based upon its damage.
This is a thread I think we probably have delved into before and it's something that will probably not go away until we revise the EF scale, even then there will be ones that fall through the cracks and get under-rated. In other words, I could list on for days until I'm blue in the face the number of tornadoes that were likely under-rated; as well there is likely a long list of ones from before the 1970s that were over-rated. The tornado itself isn't what is being rated, it is the damage that it causes and that is used to get an estimated wind speed of what caused *that* particular damage at *that* particular point.