I guess a lot depends upon how important tornado climatology is. If you value the tornado climo., then it's probably not the best idea to start mixing measured wind speeds with the F-scale; it'd probably be a better idea to formulate a new scale which can differentiate damage-assessed tornadoes from a new wind-assessed tornado. I mean, in the end, a goal would be to ensure that an F5 from the 1950s would still be considered an F5 now. A new name, such as EF, would help avoid confusion. Back to climo... Our climatological record is based on damage assessments. Now that we're finding out that the wind speed attached to the damage expected by Fujita and others in the past may be significantly in error, the tornado climo database may get royally messed up if we start letting measured wind speeds be a reason for tornado F-scale classification. Would it be better to rate each scale on measured wind speeds? IMO, most definately. However, since tornadoes have always been rated on damage production, tornado climo will become nearly useless if "wind assessments" (let's assume that's even plausible right now) become the primary mode of assigning an F-scale to each tornado... Tornadoes in the field, which would otherwise receive F0-F2 based on a lack of damage could be rated F3-F5 based on measured / determined wind speeds. On the other hand, a relatively weak tornado that moves very slowly and churns over a housing development would be rated weaker on the wind-speed scale than would the damage-scale would indicate... Obviously, scientific truth is more important than climatological continuity, but just something to be aware of!
Wind obs would be a more true representation of the 'intensity' of a tornado than damage, for reasons that are obvious and/or have been discussed before, but, we also must realize that VERY few tornadoes are even sampled at the ground (maybe the couple of ASOS/AWOS obs that have been directly hit by tornadoes, though I'm not aware of any). Radars can sample tornadoes, but the data obtained are often several hundred meters about the ground. In addition, there are inherent errors in Doppler wind velocity estimates, which, IIRC, is the main reason Dr. Wurman changed his data from 5-3-99 Bridge Creek from 318mph winds to something like 301mph +/- 17mph. Radar measurements may be a better indicator of wind speeds than an application of a damage-wind speed relationship scale, but there are still plenty of errors in radar obs. In addition, how many tornadoes are caught by Doppler radars each year? A very small fraction of all tornadoes across the US each year, that's for sure!
Photogrammetry has been applied to tornadoes to help determine wind speeds, but even those are estimates (after all, there are complicating factors such as drag and centrifuging effects). I know Tim Samaras mentioned photogrammetry being applied to his media turtles, but I'm not sure if he's done any of that yet.
So, that leaves direct measurements... Those would be nice, but the most 'controversial' tornadoes are those that are in the upper range of intensity. It'd be difficult, I believe, to ensure that an anemometer withstands F4/F5 winds. In addition, we know that wind fields in tornadoes are quite heterogeneous (e.g. not axisymmetric!). There are subvorticies, etc, which may very well 'miss' the anemometers... Then what?
Again, I do think direct or remote measurements would probably be a more accurate tool in the determination of the strength of a tornado (relative to damage / wind relationships), but both methods can (usually) only be applied to a very small number of tornadoes each year.
