Mehdi, It depends on the studied organism genome. In the case of polyploid (potato) or alloploid plants (bread wheat, Brassica napus, B. juncea, B. carinata) it is rather expected.
Apart from Alex answer, it is also possible (and rather common) to map a marker to more than one chromosome (linkage group) in diploid organisms. For recent genetic maps, made with lots of markers and mapping software, it comes down to the LOD scores. The cut-off of your LOD score is a compromise between type-I and type-II errors (i.e., to establish as linked a marker which is not, and not establishing as linked a marker which is actually linked). As in any statistical process, you can have some markers in the interface....You can solve it by getting a larger progeny or by, adding other markers which lie close to the region(s) where the problem marker is located.