Hello. I just got answer from ISC support - this behavior is correct - this is feature, not bug - so we can close this ticket. Here is answer from Mr. Shawn from ISC:
"There are two ways in which this can occur that are not problems. 1) When using failover the peers will update each other with a lease time for the a client that is in advance of what they provide to the client. This value is calculated to provide the desired lease time when the client does it's next renew. If the client renews later than expected the amount of time left in the calculation will be lease than desired and that is what will be used. 2) There is a feature dhcp-cache-threshold to limit the number of writes to the disk due to clients renewing earlier than expected. This feature allows the server to provide the client with the time remaining in the current lease if the client renews before the dhcp-cache-threshold. By default this is 25% so in the simple case if your clients are attempting to renew < 3 hours after getting a 12 hour lease they will get the remainder of the time. You can investigate this issue by checking when a client is renewing compared to when it was originally issued the lease and seeing if either of these explain the behavior."
