different results from limma using the same data set and two different methods

Hello all, I have a cDNA data set with 16 different conditions, each
with 3 pairs of dyeSwap measurements (so 6 repeats in total). I ranked
differentially expressed genes in each condition using two different
methods in Limma, and the two methods yielded different B values and
gene ranks. I do not know if this is normal or I did something wrong.
Here is how I did it:

METHOD 1: Generate a MAList for each condition, and fit the linear model
with individual MAList separately. The code looks like the following:

M.condition1<-read.delim("condition1-log2Ratio.txt", header=TRUE)
A.condition1<-read.delim("condition1-amplitude.txt", header=TRUE)
MA.condition1<-new("MAList", list(M=M.condition1, A=A.condition1))
MA.condition1$genes<- read.delim("Genes.txt", header=TRUE)

fit.condition1<-lmFit(MA.condition1, design=c(1,-1,1,-1,1,-1))
efit.condition1<-eBayes(fit.condition1)
output<-topTable(efit.condition1, number=16200, adjust="fdr")
write.table(output, file="condition1-ebayes-sep.txt", sep="\t")

METHOD 2: Generate a MAList for the entire data set (96 arrays in
total), and fit the linear model with the MAList. The code looks like
the following:

M.all<-read.delim("all-log2Ratio.txt", header=TRUE)
A.all<-read.delim("all-Amplitude.txt", header=TRUE)
MA.all<-new("MAList", list(M=M.all, A=A.all))
MA.all$genes<-read.delim("Genes.txt", header=TRUE)

treatments <- factor(c(1,1,1,1,1,1,2,2,2,2,2,2,3,3,3,3,3,3,4,4,4,4,4,4,