ObjectDB uses the ordinal Java String lexicographical order, which is based on the order of the characters in Unicode. In Unicode, ordinary English letters are located separated from other language letters.
You can change this behavior by normalizing your strings. You should store your strings in the database normalized, if possible. Normalizing the strings only during sort is also possible, but it is less efficient, of course.
See also this forum thread for more information on normalizing strings and for sample code (but notice that the code in that thread demonstrates the less efficient solution of normalizing the strings during sort).
ObjectDB - Fast Object Database for Java (JPA/JDO)