ball impact with rigid or flexible ground

I used the drop test module to simulate elastic ball impact with flat flexible surface of the same ball material properties . i meshed the whole ball with hexahedronls dyna element

at certain mesh size, good results of  3% error. but i still have the problem when i
refine the mesh gradually (for whole ball) the finte elemet curve and
hertz curve (the correct results) get closer until i reach certain refinement (which is not
too fine) ansy ls dyna curve start geting further from  hertz curve
until it coincide approximately with the 1st mesh (the coarser mesh),
what is the problem? is it the time step?or what?

this causes headache  for me since one does not know which curve
he/she must trust,  what is the criteria?, assume we do not know hertz
law then we are in trouble becase we will select the results of the
finest mesh which appeared to be not true. does finer mesh in impact (explict Dynamics) mean always better results or at least not worse than the coarser? is the mesh size connected to time step and how?