Tosafos proves from the gemara that makes the case where he leaves behind the areas between the lines, to be when the get is "me'urah". At first Tosafos understands that it refers to the letters from the upper lines (such as the lamed) actually touching the letters in the line above. Based on this we can prove that by a gett the letters don't have to be surrounded by klaf on all sides (mukav g'vil), they can even be touching one another (so long as they retain the shape of the letter). But Tosafos then says that it is not such a good proof because it may be speaking when the letters aren't actually touching, rather the lamed in the lower line goes into the heh of the upper line so that it is "me'urah" but still mukav g'vil. Even without a proof though, Tosafos concludes that we don't need mukaf g'vil by gett since it is a din in kesiva tama (as the gemara says in menachos) and that is only a requirement for STA"M. The minhag brought in shulchan aruch e.h. 125 is to be makpid l'chatchila about mukaf g'vil. The Beis Shmuel (7) explains the minhag either because we want the letters to be clear, and not attached to one another, or because we want to be makpid l'chatchila for the rules of STA"M. The nafka mina is if it the letters are touchin on the bottom which don't really ruin the form of the letter, according to the first reason it is ok, but according to the second it is not. However, the chazon ish (hilchos tefillin) understands that even the din d'oraysa by STA"M of kesiva tama and mukaf g'vil are for the purpose of having clearly defined letters. Also, based on the minhag we should be makpid for kutzo shel yud which is also a din in kesiva tama (menachos 29a), even though it is not me'akev.