Not marketing fluff - if they're using the copula based approach described above which I believe they are. Effectively it creates dynamic risk profiles based on past returns which then are incrementally adjusted based on the fund's continuing performance. The point is not to match month-to-month returns, but instead create a statistical distribution of returns that matches the targeted hedge fund over the longer term. Read the papers - even if you aren't familiar with the math (and I can't blame you, it sucks), it's fairly easy to pick up what the approach is
Also - the <b>really</b> interesting part about the approach is that you aren't limited to just cloning hedge funds. Imagine a real estate risk profile with perfect liquidity (as the quoted approach trades extremely liquid futures only). Even more interesting, imagine investigating the returns of your university's endowment fund and determining the theoretical risk profile that would have (reduced your volatility/increased your return/etc) for the last few years... a bit of math later, and you can have a custom product. This is the future.
If you beleive in replication, you might also believe in Rael clonning humans ! Seriously if I could clone a HF, I would not ! Think one second if I could clone, then I would clone a return of 100% and vol of 1%, not 14 and 8.... Second, copula or factor, it's all on historic data so it works until it doesn't !! In the menatime Goldman and Merril are rubbing your money... but I agree that factor model is real joke and copula is a smart proposal !!