I havent done any testing in a long time so am rusty, but to some extent it depends on what type of failure you're expecting. For example, if using skins that have fibres 90 degrees to the length, then skin failure could occur quickly as they become indented. Also depends on things like density of your alu core. The first flexural tests I did was about 30mm width, (something like 80mm span from memory) which was too thin to get a proper consistency with full cell volume/unit area, so the bigger your cells, the wider you make it. Also, you'll want a good 20mm radius if you can on the loaders for realy flexural tests but if you are trying to simulate more point like loads, this could be reduced - really depends what you're trying to get out of it.