Mick Hopper did a very useful paper on the matter of retention at the 2008/9 (I cant recall which) JCT Symposium.

Ive got a copy in pdf format, would permission be forthcoming to post it here?