1. I guess I did not realize sensitivity to tilt. So trying to move by hand is not a good idea. Thanks Neil.
2.Thanks to Achim, with his post I realized that I made a mistake in offset calculation ( I had ignored resolution of 15.6 mg, as pointed out by Achim).
3. As I try to find/understand offset calculations, I am learning that, it could be dependent on part/board. I see that I have two boards with ADXL345 on it - and I see that their zero-g readings are different ...Achim, you alluded to 'integrated offset calibration' - I do not know about it. Can you please point me to any such dynamic offset calculation algorithm ? Any knowledge/articles which points out viability of such procedures, would help me.