The CNet reviewer might have some strange testing parameters. He reports the 4500 as having less range than the 3800 in both 2.4 and 5 GHz bands and less throughput in the 2.4 GHz band.
(FYI - The first 4500 I had had poor 2.4 GHz and superb 5 GHz performance. The second 4500 I had was the opposite in performance. My current [third] 4500 works a treat.)
I have not seen this disparity of results for single channel, non-bonded 2.4 GHz band performance that CNet states with my current 4500. While certain mid range locations I test the routers in have both the 3800 and 4500 performing similarly, with the 3800 occasionally edging out the 4500, the 4500 has significantly better throughputs, up to ~30 % more, in long range and one mid range location I've tested it in.
To make matters more interesting, using the single channel non-bonded 2.4 GHz band, the link rates at my more remote location may be less with the 4500 BUT it has consistently higher throughputs than the 3800.
Link rates for bonded channel 5 GHz band are always higher for the 4500, and so are the throughputs. A problem, though, is that sometimes 5 GHz connection fails while testing for either router at distances, so it is difficult to say which one is the more reliable overall. Nevertheless, the 4500 typically has significantly greater throughput over all locations I test in, including remote locations.
Looks like CNet does limited location testing and/or has competing signals and/or obstacles which resulted in the confusing data.
Furthermore, CNet testing the 2.4 GHz band with a Mode of 450 Mbps produced even worse results than a Mode of 217 Mbps, telling me their location suffers from other 2.4 GHz neighborhood signals or they had a lemon.
Too many unexplained variable, such as how they test and what obstacles to signal they have.
I'm not buying the results of their tests as adequate measures of the 4500's abilities. The 3800 is a great router, but so is the 4500, though you wouldn't know it from CNet's review. I don't doubt these were the results of CNet's testing, but I do doubt their results are typical for others. This is why one cannot know for sure how a router's wireless signal will perform in their location until they try it out for themselves.
