In your code, at each iteration, if a character is found in a str1, a new string is created. It turns out that in the worst case you will get str1.length rows in memory

This code leaves punctuation intact except for spaces. If the number of spaces is important, then care should have been taken to create suitable tests.

Yep, in the tree that is described even in the example, this solution returns the incorrect value "15" instead of "-51".