If you try to do it the other way (make a directory /net/whatever/local and mount the local fs there, then start the automounter), the autofs will be "on top" of the local fs, so be hidden.
It doesn't seem to be like that, the actual directory in question that is automounted (among many others) is /proj/chris. This is mounted from a remote Solaris computer called crusade such that 'mount' reports:-
/proj/chris mounted on crusade:/scratch/chris
(which seems the wrong way around to me, but still). /scratch/chris is a 'real' file system on crusade which is exported by crusade.
Well, I've got it done the 'right' way. I looked at the "other end", i.e. the Solaris box that was sharing the filesystem and it turns out that share can specificallly exclude targets. So I got the sysadmin (who has root on the Solaris boxes) to do:-
unshare /scratch/chris share -F nfs -o rw=-chrisg:-ricky:ourdomain /scratch/chris
This means that the share is explicitly *not* available to chrisg (my Linux box) and ricky (another Linux box I use occasionally). Thus the automounter doesn't pick it up and I've got rid of the directory on the Linux boxes. I can now create my own /proj/chris on the Linux box and have what I need.
By the way it would appear that you can mount as many things as you like in one place and only the last one is left working. Before doing the above I could get the effect I wanted by mounting a local file system 'on top of' what the automounter had done. However I couln't automate this by putting the mount in /etc/fstab because autofs came along and did its thing after that.