That would explain why a Cleveland player's win shares HAVEN'T INCREASED MUCH.
But not at all why they would have DECREASED. (Right?)
I think the way a guy's WS could DECREASE would be......well, I think 2 things would be needed:
(a) That the team won just a very small number of games in the interim (i.e. what you said), PLUS.....
(b) That OTHER GUYS ON THE TEAM piled up nicer numbers in these last few games, with the result that they would get MORE CREDIT RETROACTIVELY for past games won this year.....even though most of these guys' new numbers were in games that the team LOST.
I don't know if this "phenomenon" has been discussed before....and I'm not sure it's been formally "discovered" although I imagine it must be obvious to people like Bill James.
It points up what we might call a flaw in the WS system: Guys get credit for what they do in the team's LOSSES.
If we were to calculate Win Shares in the most logical way, I think we would look only at the games that the team WINS. (Which would be harder.)
Among the advantages would be that aberrations like THIS would be less common, although I think they'd still be possible.