R lme4 predictions -

i have fitted straightforward 3 level model data.

 model.3l <- lmer(diff ~ final.mark + max.mark + (1 | unit.code) + (1 | unit.code:item), data = seed.data) 

i have created set of predictions using fixed effects model.

predicted <- predict(model.3l, newdata=null, re.form=~0) 

i have extracted parameter estimates , random effects (code not included cross checked). here comes naïve question. why predicted values lme4 differ considerably can calculate manually fixed effect parameter estimates using following code?

predicted <- seed.data$fix.intercept + (seed.data$final.mark * seed.data$fix.final.mark) + (seed.data$max.mark * seed.data$fix.max.mark) 

here more detail requested. summary of model:

linear mixed model fit reml ['lmermod'] formula: diff ~ final.mark + max.mark + (1 | unit.code) + (1 | unit.code:item)    data: seed.data  reml criterion @ convergence: 604847.5  scaled residuals:       min       1q   median       3q      max  -20.6746  -0.3056   0.0119   0.2791  17.2782   random effects:  groups         name        variance std.dev.  unit.code:item (intercept) 0.02895  0.17015   unit.code      (intercept) 0.00790  0.08888   residual                   0.45422  0.67396  number of obs: 294678, groups:  unit.code:item, 395; unit.code, 19  fixed effects:              estimate std. error t value (intercept)  0.016734   0.027614    0.61 final.mark  -0.184453   0.001045 -176.59 max.mark     0.092059   0.003905   23.57  correlation of fixed effects:            (intr) fnl.mr final.mark -0.001        max.mark   -0.557 -0.151 

and here few lines of data. quite apart expect predictions cases 7, 8, 9 , 12 same.

   unit.code item final.mark max.mark diff fix.intercept fix.final.mark fix.max.mark re.intercept re.intercept.2   predicted 6             1          8       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.20085239 7             1          7       10   -1    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126 -0.16805334 8             1          7       10   -1    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.10845857 9             1          7       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.01572996 10            1          4       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126 -0.16805334 11            1          6       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.01639952 12            1          7       10   -1    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126 -0.07565952 13            1          6       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.20085239 14            1          8       10   -2    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.01606474 15            1          5       10    0    0.01673431     -0.1844529   0.09205904    -0.173901     0.01075126  0.56908855