This isn't particularly rigorous. Like all of these studies that claim to support epigenetic memory in mice, they've started with mixed genetic background mice, so that none of the offspring are genetically identical to the parents. You can't rule out the possibility that genetic differences are responsible for this result.
This is publication bias in action. Studies that don't show epigenetic inheritance don't get published. Where are all the properly controlled experiments using inbred strains?
I am far from a geneticist (or in fact any form of scientist), but why do normal experimental controls not work in the case of this experiment?
You perform the odor exposure to half of your group and not the other half (the control group). Then you compare the results (measurement of the offspring) between control group and non-control group.
Assuming that your group size is large enough that the measured difference can be considered statistically significant, then it doesn't matter what other variances there are - you have controlled for them.
There are strains of mice which have very little genetic variability, thus excluding (or at least diminishing) other variations than the studied effect. This study apparently didn't use one of such strains, so many genes could interfere. Large enough is otherwise really large.
This is publication bias in action. Studies that don't show epigenetic inheritance don't get published. Where are all the properly controlled experiments using inbred strains?