为何优化前后的分析结果不同

我用混合设计设计了一组试验，经过优化后的组合结果和优化设计理论结果不同。
将优化组合也放进试验组，出来的结果和不放优化结果的分析差别很大，请高手指点！
一、分析结果
Regression for Mixtures: L0μi versus Fe2O3, MnO, ZnO


The following terms cannot be estimated and were removed:
Fe2O3MnOZnO*ZnO
Fe2O3ZnO2
MnO[i]ZnO2
1 / Fe2O3
1 / MnO
1 / ZnO

Estimated Regression Coefficients for L0μi (component proportions)

Term Coef SE Coef T P VIF
Fe2O3 -3.6435E+12 4.53136E+11 [i] 2.36717E+20
MnO -1.2360E+13 1.50512E+12 9.35650E+20
ZnO 2.30572E+12 3.19493E+11 1.02806E+19
Fe2O3*MnO 3.20090E+13 3.91664E+12 8.17 0.000 1.76622E+21
Fe2O3*ZnO 2.61496E+12 3.34123E+11 7.83 0.000 3.13458E+18
MnO*ZnO 9.56286E+12 1.09527E+12 8.73 0.000 1.20482E+19
Fe2O3MnOZnO -6.0362E+12 1.06079E+12 -5.69 0.000 3.15054E+18
Fe2O3MnO -1.7390E+13 2.10344E+12 -8.27 0.000 2.28809E+19
Fe2O3[i]ZnO 1.31626E+13 1.68873E+12 7.79 0.000 1.10682E+19
MnO[i]ZnO -3.5436E+12 4.96979E+11 -7.13 0.000 6.35121E+16
Fe2O3[i]Fe2O3MnO*ZnO -5.3429E+13 7.08808E+12 -7.54 0.000 3.92148E+19
Fe2O3MnOMnO*ZnO 4.28651E+13 5.47995E+12 7.82 0.000 8.39113E+18
Fe2O3[i]MnO2 2.94462E+13 3.62687E+12 8.12 0.000 3.06305E+18

S = 97.114 PRESS = 551718
R-Sq = 97.06% R-Sq(pred) = 93.39% R-Sq(adj) = 95.71%

Analysis of Variance for L0μi (component proportions)

Source DF Seq SS Adj SS Adj MS F P
Regression 12 8102086 8102086 675174 71.59 0.000
Linear 2 765237 705974 352987 37.43 0.000
Quadratic 3 2458687 2375020 791673 83.94 0.000
Special Cubic 1 783188 305373 305373 32.38 0.000
Full Cubic 3 1851487 1972861 657620 69.73 0.000
Special Quart 2 1621822 579137 289568 30.70 0.000
Full Quartic1 1 621664 621664 621664 65.92 0.000
Residual Error 26 245209 245209 9431
Total 38 8347294

Unusual Observations for L0μi

Observation Fit SE Fit Residual St Resid
25 3579.000 3236.999 56.069 342.001 4.31R
26 3049.000 3236.999 56.069 -187.999 -2.37R
二、优化结果
Fe2O3, MnO, ZnO=0.5297：0.3158：0.1545
L0μi =4907
三、实际结果
L0μi =3526