The scoring system is used for quantitative traits and i think the scoring system is very good as it helps in seeing what is the effect of a particular QTL among many QTLs for a phenotype. It has worked in most of the time except when the scoring is not performed with care or during the biological replicates where other factors other than the factor in question, for QTL analysis, influence the score. Anything wrong in scoring or genotype will definitely influence the results.

I fully agree with Kamal -- so long as care is taken in collecting the "score" data, QTLs detected with this data  can be very useful -- with co-mapping QTLs allowing you to identify factors that underly the score values, which are not always as obvious as you might think they would be.

In my opinion, the reliability and reproducibility of you data depend on the scoring method you are using. The data of some traits are more "objective", for example, the height of plant, the weight of fruit. because you can directly measure it with ruler or scale. In the same experiment condition, the reproducibility is good.But some traits may be more "subjective", for example, the color of fruit, the disease severity of leaf blight, for these kinds traits, people usually estimate visually and do not use any equipment. The value recorded by different people may have big variation, especially for the people have not a long-term training or work experiences.  Suppose you are testing 1000 wheat cultivars under leaf blight infection in field. Each wheat cultivar has 30 plants in a plot. When you recode the disease severity, you may not check every plant, instead just take look five or ten plants each row, and estimate either the average severity in a row or five to ten data points in a row. For the extreme value 0, 5%, 90%, 100%, people usually do not have big difference. But for the certain cultivars,you may call severities as 35%, 65% and 75%. your peer may call them as 30%, 70% and 80%. 

So I think if the data you have to estimate but not using equipment, the people should accept very well training before each scoring. This will lower the variation and improve the reliability.

so long we correlate our phenotypic data with the expected number of alleles a plant contains for the trait  in a population....data could be reliable and QTL position justifiable.

So long as the operational heritabilities (broad-sense heritabilities) for each of the scoring data sets are above 0.50, it should be possible to map QTLs using the scoring data (using adjusted means, BLUPs or BLUEs as appropriate) and these QTLs will be as valid as those for other measured quantitative traits having similar operational heritabilities (those of flowering time, plant height, panicle length, ... in cereals are typically above 0.8 if the trials from which they have been collected are well managed). However, if you do not use a good experimental design, you can map traits such as plot number -- which are entirely artifactual.

I would like to thank you for your question. As we know QTLs analysis depend on phenotyping and genotyping. Both of them need do in accurate.

Also, most  QTL mapping largely depends upon the heritability of the traits. The qualitative traits is some times difficult in scoring as it need more experience as it is not measured traits like disease. 

The quantitative traits is need to be measured in more accurate and have a replication and more plants per each replicate to get high LOD score. Also, to confirm your work you must repeated under to different locations or years or conditions.

Also, Genotyping need more DNA markers and cover the genome under study to get more reliable QTLs.