That doesn't mean after the translator has gotten within 95% of the norm the work is done. No. That's when the significant differences occur. A translation could be grotesque or wonderful and still fall within that range. What it means is that the main issues will have to do with acceptability within the target literary culture.

Let's look at an example, the opening of a Lorca poem:

Aquellos ojos míos de mil novecientos diez
no vieron enterrar a los muertos,
ni la feria de ceniza del que llora por la madrugada,
ni el corazón que tiembla arrinconado como un caballito de mar.

--FGL, "1910"


Belitt:

Those eyes of Nineteen-Ten, my very eyes,
saw no dead man buried,
no ashen bazaars of dawn's mourners
nor the heart, in its recess, like sea-horses, wavering.

Statman and Medina from the most recent APR

My eyes in 1910
never saw the dead being buried
or the ashen festival of a man weeping at dawn,
or the heart that trembles cornered like a seahorse.

What is the variance between those two versions, and the variance between either of them and Lorca's stanza? Belitt might not even reach the 95% standard. He's at about 92% maybe. Statman and Medina are at about 97%. (These percentages have no real quantitative meaning; I'm just using them as a way of communicating my own judgment.) The only thing that they leave out, in semantic terms is the demonstrative force of "aquellos ojos míos..." I'm not saying that 100% is ideal, either. Look how Belitt, in trying to keep both the demonstrative adjective "aquellos" and the emphatic force of the accented "míos," totally fucks up the line in rhythmic terms. In fact, he seems to be trying for a padded iambic pentameter there.

So the superiority of the Statman/Medina version is not that extra 7 percentage points of accuracy. I don't really care that Belitt switched a plural for a singular and a singular for a plural, that Statman/Medina intensify a "did not" to a "never."

No. The superiority is in a better feel for the text, better rhythm and diction. Would I have translated this the way Medina/Statman do? No. But that is because I am a different translator. My version might be closer or further from the norm, i.e. the average of linguistically competent translators.