read.table() versus scan()

I need to import a large number of simple, space-delimited text files with a few columns of data each. The one quirk is that some rows are missing data and some contain junk text at the end of each line. A typical file might look like:

a b c d
1 2 3 x
4 5 6
7 8 9 x
1 2 3 x c c
4 5 6 x
7 8 9 x

I'm trying to avoid having to pre-process the text files, as they all sit on an ftp site that I don't manage. ?My initial approach was just to read the files using a read.table() statement with the arguments flush and fill set to TRUE. For example, to import the above text file I tried:

read.table(file="ftp://ftp.example.dta", header=T, row.names=NULL, fill=T, flush=T)

However, R throws the error "more columns than column names" and won't import the file.

Interestingly, if I move the extra text "c c" from line 5 to line 6 in the data file, read.table() reads the file just fine, and ignores the "c c". ?So, my first question is, why does simply moving these data down a row solve this problem?

Next, I decided to try reading the file with the scan() function and it worked perfectly:

data.frame(scan(file="ftp://ftp.example.dta", what=list(a=0, b=0, c=0, d=""), sep=" ", skip=1, flush=T, fill=T))

I'm new to R, but as I understand it read.table() is based on the scan() function. This makes me wonder if there is an additional argument I can add to read.table() to make it import the file successfully, as scan() was able to do. ?Any help in this regard would be very much appreciated. ?I'd also really like to hear folks' perspectives on the merits of scan() versus read.table() (e.g. when is scan() the best option?).