Please Help - Very strange problem with internal String encoding

Question

I created a file in one-byte russian encoding cp1251 and declared String literal with 2 letters:
String str = "ab"; //attention! ab - two russian characters
After I got bytes from it - str.getBytes("cp1251") - it returned 2 byte's array.
Now I created a file with UTF-8 encoding with equal content and suddenly:
After I got bytes from it - str.getBytes("cp1251") - it returned 4! byte's array. Why?
I need to get a one-byte encoded arrays (cp1251 or koi8-r) but getBytes ALWAYS returns two-byte encoded arrays.
It is very strange: cp1251 is always one-byte encoded, but .getBytes("cp1251") returns two-byte on each letter. Why? Please help.

Answer 1

I did not read a string from a file. I created 2 .java files with different encodings (cp1251 and UTF-8) and compiled them, telling compiler with -Dfile.encoding=*** to read them correctly. While execution java interprets two looking equal in editor strings as different objects with different .intern() representation.
Why java consider source .java file encoding while creating internal representation of String object and creates from looking equally in editor strings two DIFFERENT Unicode representations. And it is impossible to convert one representation to other - impossible to get two equal byte[] arrays.