In my experience the CacheRead and CacheHits property values are always (effectively) the same, even when the cache has been flushed, and this has been true for all versions extending from 12 back to (at least) 8. Here's a test in V12...

CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
CALL sa_flush_cache();
SELECT COUNT(*) FROM SYSCOLUMN CROSS JOIN SYSOBJECT;
SELECT DB_PROPERTY ( 'CacheRead' ), DB_PROPERTY ( 'CacheHits' );

DB_PROPERTY('CacheRead'),DB_PROPERTY('CacheHits')
'448796','448539'

Clearly, either I'm as dumb as a sack of rocks, or something isn't doing it's job.

Or both.

According to the Help, CacheRead and CacheHits properties are important http://dcx.sybase.com/index.html#1200en/dbusage/perform-s-5421283.html

Detecting whether memory is a limiting factor To detect whether memory (buffer pool size) is a limiting factor, check the CacheHits and CacheReads database statistics. If these statistics are not present on the graph, click the Add Statistics button and select CacheHits And CacheReads. If CacheHits is less than 10% of CacheReads, this indicates that the buffer pool is too small. If the ratio is in the range of 10-70%, this may indicate that the buffer pool is too small—you should try increasing the cache size for the database server. If the ratio is above 70%, the cache size is likely adequate. Note that this strategy only applies while the database server is running at a steady-state—that is, it is servicing a typical workload and has not just been started.