Measuring spheres accurately

Measuring spheres accurately

When probing the BB's shown in the pictures (taking the features as a sphere to locate the x,y,z locations for each), I'm attacking it straight down using A0B0 and I'm noticing some sliding of the probe tip for each hit (cosine error?). I'm wondering if there's a "best practice" like using an angled probe tip instead at like A90B90, or will the software already compensate for the sliding I'm seeing?