Tesseract confused to identify the already trained character

I have done the training as specified in the site for burmese language.
Instead of using another scanned page, i am trying to use the same image which 
i used for training tesseract.
So this procedure should give maximum accuracy.

What steps will reproduce the problem?
1. Please find attached the trained data and the tiff file  i used for training
   (For testing i used paper scan tiff image of dpi 300)
2. RUn tesseract for the same image with the attached trained data.
3. Still the tesseract get confused with the characters. Accuracy is only 60%

What is the expected output? What do you see instead?
Since the same training image is used for recognition, the accuracy must be 
high.
I am not sure why tesseract has problem to identify the characters.
Please help me , how to proceed with this

What version of the product are you using? On what operating system?
Tesseract 3.02 on windows 7 64 bit

Original issue reported on code.google.com by manickamsp@gmail.com on 10 Jul 2013 at 5:43

rlightner / tesseractdotnet

Tesseract confused to identify the already trained character #27