This is interesting. Now I realize that I've made a major imprecision, betraying my own line of thinking.
There are two different themes here: one is the historical point of view, that is, trying to understand how original performances were. That makes us close to the composer's intention, or original intention, or original facts.
Having said that, one must recognize though that no human being is 100 % certain about his creations, and quite often composers like other's interpretations more than their owns. Mahler himself was very insecure about his compositions, and made changes during the years: changed tempi, movement orders, etc. That is a sign of a pursue of perfection that deviates clearly from the classical and romantic periods were pieces were rarely revised.
A different theme however is the quest for understanding what has been written, the score, alongside with its historical context, etc. Something that can only be done after some time. The goal of this line is to extract the maximum out of a score, interpreting what has been written. My opinion is that the best performances of a piece come from this approach, rather than trying to reproduce the original. Best in the sense that they are more informative and insightful, revealing more about the music and his author. In this category I find Klemperer's and Abbado's performances inferior to those Rattle or Tilson Thomas, and specially of Boulez's what I think is the most revealing and expressive one. In relation to the original performances however, the story is possibly the opposite, as the last posts highlight. ANd that is my mistake before, confusing the best interpretation with a piece with the composers intention. Those are simply two different things.
Well, anyway, thanks guys for your thoughts.