1. Yes, you should change that expression in your 3dcalc command.

As it is, the voxels near the edge of the brain will get a mean of
0, 33.3, 66.6 or 100 (rather than each voxel having a mean of just
0 or 100). This will only affect a few voxels at the edge of the
brain though, and will probably not have an effect on your small
precuneus results.

Open your Run_all_mean+orig dataset, and check whether there are
values between 0 and 1 (exclusively). Or you can use 3dhistog to
check it out.

2. The only thing that would cause a realtive scaling is if you had
multiplied one in (such as via Run_all_mean+orig). But I do not
think that happened.

Just to be sure, what is the output of:

3dhistog -nbin 10 Run_all_mean+orig

If you really want to be sure, go back to orig space and review
the results of each pre-processing step for at least a couple of
subjects. Make sure the numbers look like what you expect.

In the scaled dataseta, for instance, every non-zero mean had
better be approximately 100.

- rick