I think it is because taking the knight would result in losing control of the center.
Why doesn't White take the Knight?

Not being a premium member, I can't use the game explorer to see the stats, but I have to assume that the position is wrong. Usually, when white plays e4 to attack that knight on d5, black responds with Nxc3.
--Fromper

It looks like the 1600 number is incorrect. If you click on show positions it only comes up with 3, which makes much more sense.

This seems to be some bug in game explorer as when I run through the moves after 5 e4 it gives only 3 games with 5....Bg7 , which loses the N ofcourse to 6 exd5 but then when I play Bg7 it suddenly shows over 6000 games with 6 Nf3 and 1 with 6 Bc4 and none with 6 exd5 . This is obviously some problem/mistake with game explorer , ofcourse black should play 5.... Nxc3 6 bxc3 and then 6...Bg7.
it must be an error in the opening explorer....my rybka opening book has every move played by black in the previous ply as ...Nxc3 and not ...Bg7 :)

No, the move 6. Nf3 has NOT been made in that position in over 1600 games, but rather if 6. Nf3 is played the game would transpose to those 1600 games.

No, the move 6. Nf3 has NOT been made in that position in over 1600 games, but rather if 6. Nf3 is played the game would transpose to those 1600 games.
Is it actually the case that this move order is unheard of in the English opening, so no one ever gets to this line to play it? But it transposes to the Gruenfeld Defense exchange variation (totally different move order, but it does get to the transposing postion with no pawn on e4!)? So White's sixth move in that variation is 6.e4 (kicking the N) ... and from our problem position 6.Nf3 transposes to that line of the Gruenfeld exchange variation -- with black to move and I believe the mainline is 6....Nxc3! wow.
This seems quite freaky to me... (the transpostion isn't what's freaky...) it's that no one in the db ever got to this position... at some point the stats are entirely transpositional... this move order has been constructed theoretically out of transpostions... but it has never actually been played in the db! (because in this move order folks would take the N for god's sake!)... anyone following me, cuz I'm not sure I can follow myself here!?

There is something wrong with the game explorer. It notices that although the move order 5 e4 Bg7 was never played by any master (trust me, then exd5 would always be played) it sees 6 Nf3 transpose to the same position, which is the one it's looking at when it says Nf3 has been played so many times and so treats it the same as the normal move order, which I don't think it should.
In this position, http://www.chess.com/explorer/index.html?id=2021918&ply=10&black=0, why does White not take exd5? It seems the obvious move but in over 1600 games, nobody has made the capture. There must be some killer continuation but I can't see it.