I expect that the way it works (nature, that is) is that in general tetraploid seeds are larger than diploid seeds. That would not mean that all tetraploid seeds are larger than all diploid seeds. What it would normally mean is that if one were to measure the size (or weight) of the average seed of a substantial number of diploid cultivars and of a substantial number of tetraploid cultivars then the average diploid seed size would be smaller than the average tetraploid size. The two average sizes would be different enough that one could state that they were significantly different but not necessarily that there was no overlap in sizes.