Gaussian convoluted with an exponential using erfc function

I have the below function with the below variable values but it contains a kink at x=0 (the maximum) and at x=-20 (the beginning of the steep increase in gradient).

y2=(h*o/t)*sqrt(pi/2)*exp(0.5*((o/t)^2)-((x2-m)/t)).*erfc(1/(sqrt(2))*((o/t)-((x2-m)/o)));
o=0.009931621960609; h=170; t=130; m=-10;

t is the lifetime and o is the standard deviation of the data so i don believe these can be changed, the other variables changing does not seem to smooth the function. attached are images of my graph (Exgaussian graph) and of the ideal shape i am aiming for (fig3).

How would i go about iterating to this desired curve shape?

[John D'Errico]

The expression you show for y2 has an obvious singularity at t == 0. Nothing you do to play with the parameters of that functional form will change that.

Does that expression truly represent the convolution of a Gaussian with an exponential? I wonder if that is true. You would need to show how you derived that expression, because it matters not how you try to manipulate that expression, it has this fundamental characteristic, yet your data seems not to.

Too lazy this time of day to actually think about mathematics, so I allowed Google to think for me.

https://math.stackexchange.com/questions/685041/convolute-exponential-with-a-gaussian/872561#872561

And what I see is an expression that seems similar to yours, yet fails to have a singularity at 0. It does make me wonder if just possibly you have made a mistake in your code or in the expression you are using.