Sir, to correctly pin point the source of this error specific information regarding the problem is required.

• Are you processing multiple channels of data or just a single channel of data. because if you are processing multi-channel data with the EMD separately and then looking at the modes the EMD does not guarantee that the corresponding IMFs of the different channels would contain the same mode. 
• Are the modes that you are interested in analyzing, very close in frequency domain i.e.  Suppose Fa and F b are modes with two different frequencies a and b.  Then the separation of the modes is strongly dependent on the ratio of the amplitude of the two modes. This phenomena is discussed in adequate detail in one of the links attached (Sorry for inconvenience I do not know how to rename the link texts ).

As per my understanding both of these problems can cause mode-mixing i.e. representing the same mode as two different IMFs or two different modes in the same IMF. These problems occur as the EMD is a heuristic method with no strict mathematical definition. The EMD is also effected by the sampling and if the number of extremas is not adequate the method may not produce optimal results. This solved by using the noise assisted EMD (NA-EMD).   

The mode mixing problem, multi-channel processing issue and the sampling issue are solved in the MEMD (Multivariate EMD). This algorithm has both the NA versions and can also be used to implement the single channel EMD. I am providing links to some works below for assistance.

https://hal.archives-ouvertes.fr/ensl-00113834/document

rspa.royalsocietypublishing.org/content/466/2117/1291

www.commsp.ee.ic.ac.uk/.../EMD.../NR_CP_NH_DM_EMD_via_MEM.

Dear Muhammad Zubair Ahmad and Ifigeneia Antoniadou, thank you very much for your kind answers. I am reading the documentation provided by Muhammad.

Sincerely,

Rocco