The simple answer seems to be that the QGD
isn't actually scoring that poorly.
Where are you getting those percentages?
I'm not sure what your database is like, but by using Bigbase 2006 (updated with TWIC every week) Black appears fine.
Not as good as in the Nimzo, but certainly not 65%-70% for White in the standard lines.
After 1.d4 d5 2.c4 e6 3.Nc3 Ng6 4.Nf3 Be7 5.Bg5 (or whatever move order you like to reach here):
5...h6 gives White 54% (a.k.a. 'standard').
In particular, the Tartakower scores better than standard (as you mentioned), but even the much maligned Lasker's yields only 57% which is far far away from 65-70%.
The Orthodox/Capablanca lines (i.e. lines with ...Nbd7 and no ...h6) all yield around 60% historically.
During the past 5 years the number of draws is increasing and white wins are down
From the usual 5.Bg5, I show White scoring 59% (and after 5...h6 white scores 57%).
5.Bf4 scores a hefty 65% ... however, after a quick ...c5 the percentages are all down under 60%
The lines with Bxf6 are 55% historically but have been scoring better in the past few years (mostly due to rating discrepancy).
Alatorsev (3...Be7) is giving 58% in the past 5 years, etc. etc.
Anyway, I hope this helps. Of course, never forget that statistics lie.
The truth is that the QGD is super solid, has a terrific pedigree, and is in no danger of giving up 65%-70% to White in its mainlines.
Don't believe the Hype.
Now go play the Nimzo and enjoy nearly 51%.
Cheers.