because the text is normalized with toUpper and uppercase
ß is SS and using the insertion index of the normalized
string in the original string is out by one

if we normalize with toLower then according to
icu/source/data/unidata/SpecialCasing.txt
then as long as the language is not Lithuanian
its always one to one

so just normalize to lower case with blank locale

Change-Id: I0ad1c7e2ad0216b7c27ec62105f9a25672ade8a8