The coefficients from the multiple linear regression are correct. You cannot just add all of the individual trend lines to get the overall trend line. The multiple linear regression attempts to minimize the error term while accounting for all of the variables.
An example of this is if you were to gather data on riding your bike. Let X1 = gear, X2 = pedal speed, and Y = bike travel speed. We know pedal speed is directly related to travel speed. However, when you upshift gears you don't pedal as fast.
If you were to look at X2 vs Y then it would show that pedaling faster decreases travel speed. This is why we must use multiple linear regression on all the variables that affect Y to get their true coefficients.