Some days ago I played six money games on TMG, with beavers enabled.  In five of the games, my opponent doubled me at first opportunity, and in four of those I beavered.  Afterwards I imported the games into GNUbg for evaluations, and could see that, not surprisingly, my opponents doubles were big errors and that the correct response on all five were beavers.  In the six games, I was happy to see that I had only committed two small cube errors.  In the 'session analysis' window, I was thus surprised to see that GNUbg had evaluated my cube error rate to be over 100, in 15 close or actual cube decisions!  When I add up my cube errors manually, I get to a very different number than GNUbg's total cube error for me.  It seems that my beavers were treated as doubles, as if I also doubled in the close-to-opening positions.  I attach the *.tmg file so you can look at it yourselves.  I hope you will fix this bug, and also make full beaver analysis available.