I am trying to fit some data but getting the wrong coeffs!

So I am trying to find some coeffs with Matlab.

Here is my code:

zdata = [2,8,12,20,26,40,50,59,68,79,82,92]';
ndata = [2,8,14,20,26,50,74,76,94,118,126,143]';
adata= [zdata+ndata];
bdata=[28.3,127.6,216.7,342.0,447.7,783.9,1050.0,1124.4,1320.7,1559.4,1636.4,1783.9]';

%define av term
av = adata;

%define as term
as = -adata.^(2/3);

%Define ac term
ac = -[(zdata.^2-zdata).*adata.^(-1/3)];

%Define aa term
aa = -[(adata-2*zdata).^2.*adata.^(-1)];

%Define ap term
ap = [adata(1).^(-1/2),adata(2).^(-1/2),adata(3).^(-1/2),adata(4).^(-1/2),adata(5).^(-1/2),adata(6).^(-1/2),adata(7).^(-1/2),0,adata(9).^(-1/2),0,adata(11).^(-1/2),0]';

X= [av as ac aa ap];

a = X\bdata

When run it outputs:

a =
  16.331878598287652
  19.691560324192896
   0.768989561715635
  23.162609278749475
  27.946210266012489

The real values of the coeffs are

15.85
18.34
0.71
23.21
12.00

So not so far off, except the last one!

Is there a better way of doing this?

following the code above I am trying to fit my data to

bdata=a(1)*av+a(2)*as+a(3)*ac+a(4)*aa+a(5)*ap

Where av, as, ac, aa, ap are defined.

How can I do this using least square fit? Thanks!