[001sky]

As other commeters have mentioned, you're ignoring what is possibly good about the data. For example, if you are trying to study a perforance envelope you may want to omit samples with known flaws. Such a sample might not be representative, but it may provide insight into potential | obvious flaws or weaknessed not obeserved. So, from a bayesian perspective it might be interesting. Also, there is the other tack of abstracting out the culture and focusing on the science (biology). Thirdly, the strategy of using this data as a single composite to be compared to similar comoposite data (eg, stratified sampling) down the road. This is all without even evaluating the data present, but it is just some reasons why its premature to "jump the gun" in the way you suggest. Better to drop the pre-conceived biases and keep an open mind until something better comes along. You have to remember that a 75 year data set is a rare thing as a historical fact, and thus dismissing it due to the limitations of the historical period (which are inseperable from the data) is throwing the baby out with the bath. {etc}