Sequence reads are always from 5' to 3', but the forward and reversed reactions are running along the opposite strands. The 2 complementary DNA strands are oriented in opposite orientation, and sequence reads from either end are generating results of those 2 different strands. See this page here - it shows DNA replication, but that follows the same logic as DNA sequencing:

http://cnx.org/content/m45475/latest/

I don't think this is the case for SOLiD paired-end sequencing. Adaptors P1 and P2 are ligated to both ends of the DNA molecules, and the DNAs are amplified by PCR. After the DNAs are denatured into single strands, the P1 Adaptors bind to beads in the orientation of 5’ to 3’, and a 3’ end modification to the P2 adaptors allow the beads to stick to the slide. In my case, the "F3 Tag" was the first read in the pair, and it was 50 bases long. It is always sequenced 5’ to 3’. On the other hand, the "F5-BC" Tag was the second read in the pair, and it was 35 bases long. So, I think it is sequenced 3’ to 5’. I need to verify this but am having trouble finding documentation. Sequencing methodologies are difficult to understand. If I find out that I am wrong, I will probably delete this message.

If you are trying to decipher an output SAM file, it can be extremely confusing when trying to understand what SAM flags are, figuring out whether or not a read is properly paired, and dealing with coordinates and insert lengths. Once you know what you're doing, you can indentify midpoints and even know what strand the paired read occurred. If you need help with dealing with a SAM output, I can assist as I went through the pain of learning it on my own as documentation is difficult to find or seemingly nonexistent.