what may have gone wrong? - (Jun/05/2018 )

Hmm- I agree with trof: that is a dodgy looking gel in Fig. 5c; there is so little separation between the bands that you could account for the change from left to right by saying the gel was smiling. Also, my problem with this is that they are trying to make quantitative judgements at an endpoint. It is quite possible for both spliced and un-spliced cDNAs to amplify to plateau in an endpoint PCR no matter how different the amount of starting material.  If it was me I‘d design 2 sets of primers to distinguish the two forms, and do qPCR.

I had done end point PCR to a cDNA for the purpose to detect the loss of imprinting of a gene.  When I received the job to continue that project the first thing that I did was changing the primers so the result of the enzyme cleavage would be an asymmetric cut with a very distinctive  pattern that there was no doubt about the results..  After some time was able to convince the boss that we can do a better assay with a QPCR and I designed 2 probes each with a different fluorophores to detect the cut/uncut template. 

If possible change the assay to a QPCR either designing 2 primer set that can distinguish between both forms as OldCloner told, good part sybr green is cheaper downside you have to assays for each primer set apart.    The other way to do the QPCR is using one primer set and 2 probes (make sure fluorophores don't overlap), the good part of using  probes its that no matter if you have either form or both forms at the same time you will be able to detect it without doubts in a single reaction so you will save in enzyme and time. QPCR may be a bit pricier, but much pricier is repeating assays that don't give a clear result and  more pricier is a rejected paper because results aren't clear for the reviewers.