AFAIK update rate is one per second.
Nope, definitely twice per second. But even if you only take every other transmission, you would still have more than enough data points to produce a pretty decent acceleration curve.
Don't think ADS-B is a solution for improper performance calculations. ADS-B relies on GS (time delay!), performance on airspeed.
I think you're missing the point. Of course the wind on the nose affects the relationship between groundspeed and airspeed, and hence aerodynamic performance.
But that's not what we're talking about here. Instead, we're using Newton, rather than Bernoulli - i.e. initial takeoff roll acceleration as a proxy to deduce aircraft mass (for a given assumed thrust).