My blog has a QuickTip on this very topic. Check it out in my sig.
In effect, there's two things to consider with IC2 wiring: Packet Size, and Rate of Loss.
With Glass Fibre Cable, your maximum packet size is 512. You can have many smaller packets running through it, even if the total exceeds 512/t, but the largest without burning out the cable is 512.
Rate of Loss is *PER PACKET*. This is where higher voltage sees a significant rate of return, because the individual packets can be so much larger than a one or two EU ding is scarcely noticable. For Glass Fibre Cable, EU loss is 1 per 40 blocks.
Therefore, if you had 100 solar panels in a solar flower configuration, wired up with glass fibre cable running 41 blocks to a BattBox, you would end up with 0 energy into the battbox, because each solar panel generates 1 EU/t, and when it passes 40 blocks, that 1 is lost for every one of those 100 packets. However, if you had them feed into a batbox before you hit the 40 block mark, THEN went 41 blocks to an MFE, you would lose far less. Each packet is 32 for LV which the battbox outputs. So that would be 4 packets (three full ones at 32, and one tiny packet at 4 EU), so the EU loss would only be 4. If they all went directly to an MFE before 40 blocks, then it would output MV or a max of 128 which then went 41 blocks to an MFSU, and you would only experience a 1 point loss out of your 100 energy.