The 2.x version of Electrum uses mnemonic sentences that are 13 words long. However, 1.x versions use a sentence only 12 words long.
I realize that 2.x changed the derivation of key pairs significantly, but I'm not clear why a sentence of 12 words wouldn't work just as well as one of 13 words.
Why is an extra word now used?
The seed phrases in 2.0 are longer because 8 extra bits are required in order to store the seed version number. (default entropy is 128 bits)
Note that the number of words in the seed phrase depends on the size of the wordlist used in the generation. Currently, the English wordlist used in Electrum has 2048 words, which results in 13 words per seed phrase. This wordlist can be extended in the future, for example 2581 words would give you 12 words per seed.
