The so-called "left" and "right" border primers both are supposed to face outwards into the genomic region where the TDNA insert landed. Most folks use a gene-specific primer (F primer or R primer that you design) paired with a left border primer because the left border of the TDNA is more likely to be present due to the way the TDNA inserts into the genome.
F primer LB RB R primer
[gene of interest][gene of interest]
The assumption is the the TDNA insertion is so large that you can't make a product from F to R. You use the estimated location of the TDNA from the stock center as a guide for designing these primers. General rule, see where they say the insert is, back off ~500 bp, and look for a good primer.
So, you set up PCR with F + LB; F + RB; R + LB; R + RB to see which way the insertion is facing in the gene of interest.
Let's say the F primer and LB give a good strong band in your mutant allele. The F + R give a strong band in the wild type allele.
Set up 2 PCR reactions for each plant you want to genotype.
Homozygous wildtype: band from F + R; no band F + LB
Heterozygous: band from F + R; band from F + LB
Homozygous mutant: no band from F + R; band from F + LB
Include a FULL set of controls and double-check any that you think are homozygous (PCR can fail for lots of reasons).
In reality, TDNA insertions are complicated, often contain insertions or deletions or multiple TDNAs or chromosomal translocations (or some combination of all of those things). That was the basis of my thesis.
Assuming LP and RP are the gene-specific primers flanking the insertion (e.g. as in http://signal.salk.edu/tdnaprimers.2.html), the PCR reaction does not usually yield a band because the duration of the extension phase (at 72C) is set to a value that is long enough to amplify the wild-type allele (i.e. LP and RP are often located about 900 bp apart) but not long enough to amplify the mutant allele (i.e. carrying the insertion).