Updated: see Ryan Fisher's answer to this question, SP800-52 revision 1 was later released, and it's double the size of the original 2005 version. It was withdrawn because it lacked information on contemporary TLS versions and known security issues (i.e to prevent misinforming, until it could be updated).
It was not "retired" (or "expired"), it was "withdrawn", admittedly a minor semantic distinction. NIST have the following generality to say about that:
This page contains a list of withdrawn Special Publications (SPs) that have either been superseded by an updated SP or is no longer being supported and no updated version was released.
is was no (evident) successor, it may be no longer supported. It's worth noting that its raison d'être is now no less than 17 years old.
I suggest the main motivation is the (selective) elimination of one or more of 3DES/TDEA, MD5 (which NIST have long disliked) or SHA-1. The official statement is 3DES is good until 2015 (or 2030), TLSv1.1 drops the mandatory TLSv1.0 cipher TLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA in favour of TLS_RSA_WITH_3DES_EDE_CBC_SHA, and TLSv1.2 drops that in favour of TLS_RSA_WITH_AES_128_CBC_SHA.
TLS versions after TLSv1.0 still support MD5, but since SHA-1 is deprecated, and must go away for certain purposes this year, there are clear advantages to TLSv1.1 and later including less dependence on MD5/SHA-1 and better support for arbitrary cipher or hash functions by way of extensions.
(So, what CodesInChaos said, but with more handwaving, conjecture and hyperlinks ;-)