I tested 2100T too and never got such a good result (only a few watts lower than 2100 at idle, 10-15W lower at load). This agrees with the results by several 2100T users. On the other hand, some users reported much better results (like you).
Here are actual measurements of the power consumption of a 2100T system at idle by Japanese users (the AC voltage is 100V so that the efficiency of PSU is worse than 120V/225V).
- ASRock H67M-ITX, 2 x 2GB, 1 x 3.5" HDD, pico-PSU150T: 26W
- BIOSTAR H61MH, 2 x 2GB, 300W 80 PLUS GOLD: 23W
- Foxconn H67S-B3, 2 x 2GB, 1 x 2.5" HDD, pico-PSU120: 20W
- ASRock H67M-ITX, 2 x 2GB, 1 x SSD, IN WIN IP-AD160-2 160W: 28W
- Jetway MIH67M-HU3, 2 x 2GB, 1 x SSD, TX850W: 36W
- ASRock H67DE3 2 x 1GB, 1 x SSD, PSU?: 46W
- ASUS P8H67-M EVO, 2 x 2GB 1.35V, 1 x 3.5" HDD, 750W 80 PLUS Silver: 39W
So the result depends on lots of factors (motherboard, HDD, PSU, and maybe the chip itself). On the other hand the power consumption at load is uniformly much higher, close to that of 2100 (one shouldn't expect 65W-35W = 30W difference).