Reporting looks for a wrong ancestor

While looking at benchmark results from
http://cbuild.validation.linaro.org/benchmarks/gcc-linaro-4.7%2bbzr115048~christophe-lyon~smin-umin-idiom/logs/armv7l-precise-cbuild446-tcpanda11-cortexa9hfr1/
I noticed that there is no *-diff.txt file.

It turns out that the reporter script is looking for a reference in gcc-linaro+bzr115047 which does not exist.

This information probably comes from
snapshots/gcc-linaro-4.7+bzr115048~christophe-lyon~smin-umin-idiom.txt
and maybe findancestor.find should actually check that the filename it returns is actually valid?

All other methods in findancestor seem to return '' in this case (find_via_variant, find_via_build and find_via_release).

I'd think that it should choose gcc-linaro-4.7+bzr115034 which does contain valid results.

Or, if we think that it shouldn't use that rev as reference, the user should get a warning about this, to help him debug such a situation.