ADE7868A Neutral Current Mismatch

Hello, everyone!
We use ADE7868A for 4-element 4-wire electricity meter. We need to detect neutral current mismatch in this meter, so we expect that we will use Neutral Current Mismatch function of ADE7868A. 

We made a PCB, it was not big deal, as we used in our previous design ADE7858A. We added 4th CT the same, as for each phase.

Maximum current is 100 A, Nominal - 5A, Voltage is 220/380V.

It seemed to be an easy task, so we changed our firmware for host to handle MISMTCH event, and added an ability to configure ISUMLVL threshold and began to test.

And we faced the problem. 

It is required to configure minimum current difference 0,5 A and higher with some step, e.g 0.5 A. (0,5 A - 1 A - 1,5 A and so on).

And during test we noticed strange behavior of Neutral Current Mismatch mechanism.

If  (each) phase current is up to 10 Amps everything works as described. If we add tampering or generate disbalance - function works.

But, if currents are higher we found that ADE7868A fires MISMTCH event, even if currents is OK.

For example, if we have I_PHA = 40 A, I_PHB = 0, I_PHC = 0, we expect i_N = 40 A also. But, if we set ISUMLVL = 0,5A we will always receive MISMTCH event.

For maximum current we need to set ISUMLVL >= 5 A to avoid wrong events.

We have lab equipment to generate currents and we checked generators, and we really have big, but the same currents, but ADE fires MISMTCH.

Power measurement is fine, and in normal conditions 0-0,3% of error.

Or we are doing something wrong, or we don't understand how it should work.

Any ideas? Is this behavior OK?