I think the calculator is sometimes inexact because it might have been programmed with a flaw in its system of processing big decimal calculations.
All calcuators are only correct to a specific number of decimal points.
Then rounding errers can always multiply as more calculations are done on them.
Then solution is,
3.1496247821038495…- π = 0.008032128514056261537356616720497115802830600625… or nearly = 2/249
The other answer might be also that the calculator sometimes rounds up or down in a correct way but when it comes to x^n it will have slightly solved for different decimal outcomes by the potentiality of each values interchanging in the calculation operation through combined multiplication and division which might be a good topic for the problem about the inaccuracy in the subject of computer science.
I took two huge numbers in a ratio that had become an approximation to near at the constant of Pi. Then I subtracted the original Pi from the approximation of Pi to receive the remainder of the approximation of Pi which is nearly the fractal 2/251 and starts with 0.0080321285140562 until the 16th decimal places of the remainder. The last decimal places are continuously wrong results given by the calculator because they don't match the correct solution by the upright integer sequence. I wanted to show you that Pi can be calculated perfectly precise and accurate but the calculator is not constantly matching it. If you like you can check it and proof the findings because it took me by surprise that three different kinds of decimals could have come up just because of the flaw from inexactness which the calculator solved me for but in turn, it gave me a great alternative solution for the research on the circle factor which is a significant mathematical methodology.