Different results from Linest and Trendline (graph) from a 6th degree polynomial

Hello,

I have some measurements, which I want to curve fit using a 6th degree polynomial.

I've done this using the Trendline from a XY Scatter graph and also using =LINEST(yvalues;xvalues^{1,2,3,4,5,6}).

The results are very different and I don't understand why.

Trendline: y = 722.55x6 - 24063x5 + 332574x4 - 2E+06x3 + 1E+07x2 - 2E+07x + 2E+07 ; R² = 0.9904

Linest: y = -35.877x6 + 880.92x5 - 8402.9x4 + 37794x3 - 71321x2 - 0.0009x + 118396 ; R² = 1

To evaluate which is most accurate, I put my x-values in the above equations and compared with test data.

"Linest" is very close to the graph from the datapoints.

"Trendline" is way off, but the equations are almost identical. Coefficient c4 changed from 332574 to 332577.

Hopefully I've done some simple mistake, but I cannot find it. Would really appreciate if someone else did. I'm not sure how to include the Excel-file, but at least I could include an image below. 

Regards,

Albin