Whole genome amplification protocols

Probably the second most asked question that I get is with regards to quantity needed for 454 pyrosequencing. I always feel like the brat visiting Santa asking for too much. Truth be told, 3-5 ug is what the Roche protocol calls for. My inside source says 1 ug is the absolute minimum. The problem isn't that you need that much material for the single-stranded library. It's how you get to the library. Namely, using Qiagen columns which lose ~50% of the sample for each of the 3 purifications.
What do you do to make more sample from what may be all you will ever get? There are two main options: 1) whole genome amplification and 2) PCR amplification using the sequences of the Roche adapters. The first option can be achieved through commerically available kits such as Genomiphi from GE Healthcare or Qiagen's equivalent.
The second option has been formalized in a recent paper by Blow et al in Genome Research. We have used the Blow et al protocol for amplifying bacterial genomes, euk cDNA, bacterial cDNA, and metagenomes without any discernable bias (also noted by Blow et al). We have done some pilot projects with the Genomiphi and the jury is still out. We get lots of product but see a higher level of shortQuality reads on our test 454 runs. We have performed several runs where the PI prepared the Genomiphi sample and they look just as good as a normal library (i.e., ~100 Mb on a FLX run).