"maximum overhead can be 1 -2 microsecond
I am getting overhead of 4 microseconds
"

That is assuming your crystal is clocking *EXACTLY* at 12.000 MHz. How sure are you about this ? Check out the ALE which should be at 2 MHz.

Best is to switch to a derivative that can use T2. No need to reload. Or another derivative with PCA will be even better.

Or a quick method would be to adjust the current re-load THAT WILL PRODUCE the frequency you want.

Raghu