In Linux, how does one system adopt two policies for thread priority?

In Linux, how does one system adopt two policies for thread priority? -- Linux general technology-Linux technology and application information. For details, refer to the following section. Daniel P. Bovet, the third edition of Linux kernel, says: Real-Time threads are 1-99, with 1 level reaching 99.
However, another argument is that the kernel uses the priority of more than 1000 threads as the priority value during scheduling-which corresponds to the lowest level 1 (Baidu article "Linux Thread Scheduling", three Scheduling Policies for Linux kernel)
The actual test is as follows: create two real-time threads 90 and 91, where 90 is in the while endless loop, and 91 is then in the while LOOP, sleep for Ms before running, level 91 threads can seize level 90. why.
The smaller the number of a common thread, the lower the priority. How can a system adopt two policies?