First of all, WHY Strings and not just numbers? For a 7 KYU this constraint just causes way too much problems with padding 0's at the end (At least when working in Haskell).
Second of all, the rounding and float imprecision cause additional problems. I have 5/10 Tests passed, with two tests not passed because of the missing end zero and the other because of some weird imprecision (eg. I get 2.39, actual should be 2.40 or 2.38).

It could be debatable if the description didn't state neatly the contrary. Your suggestion is not a suggestion (what do you actually suggest?), and if the idea was to change the kata design, it won't be done since it would invalidate more than 50,000 submitted solutions.

This may be debatable, but if the ball is starting at the same height as the window the result should be 1 and not -1

edit: especially since there's a possibility that the resulting height after a bounce is the same as the height of the windows, thus falsifying the result