awk is possibly not the best tool to use here. Since it's designed for breaking things up into fields, manipulation of the delimiters between the fields is sometimes not so easy to do, particularly when you only want to affect some of them.
I'd use sed in this case, since it targets the line as a whole.
sed -r 's/(^[^[:blank:]]+[[:blank:]]+[^[:blank:]]+)[[:blank:]]+/\1/' infile.txt
This assumes that each field is separated by whitespace. It's a simple regex that matches [linestart][nospaces][whitespace][nospaces][whitespace], and prints all but the last bit of whitespace back into the line, effectively removing that part.
There may be easier ways, but it's the first solution that came to mind.
Edit: D'oh! Just after posting I realized that there is a much easier way.
sed -r 's/[[:blank:]]+//2' infile.txt
The two at the end means the substitution only affects the second match on the line. So we just tell it to match contiguous spans of whitespace, and substitute the second match with nothing.