Beedtools intersect: more peaks intersected than intersected portions of those peaks?

Question

[salamandra]

When finding the number of peaks/genomic regions in A that intersects with B and C (note even the non-overlaping part of each peak is reported):

 bedtools intersect -f 0.5 -r -wa -a $A -b $B -sorted | sort -k1,1 -k2,2n | uniq > Apeaks.intersectingwith.B 
 bedtools intersect -f 0.5 -r -wa -a Apeaks.intersectingwith.B -b $C -sorted | sort -k1,1 -k2,2n | uniq > Apeaks.intersectingwith.B.and.C
 wc -l Apeaks.intersectingwith.B.and.C

we get 7 peaks/regions

But, when finding the number of regions that overlap in A, B and C (only the overlaping part of each peak is reported):

bedtools intersect -f 0.5 -r -a A.area.intersecting.B -b $C -sorted | sort -k1,1 -k2,2n | uniq > A.area.intersecting.B.and.C 
wc -l A.area.intersecting.B.and.C

we get 6 regions

How is it possible to have more peaks that intersect (in A), than intersecting areas of those peaks?? I don't get it..