It is said many words here http://www.unicode.org/reports/tr38/#SCTC about kSimplifiedVariant and kTraditionalVariant field cases (5 cases). But I didn't understand, if these two relations reflect each other? For example if A is simplified version of B, then does it mean that B is traditional version of A?
Let me summarize the cases and try to answer your question (TC = Traditional Chinese character set, SC = Simplified Chinese character set).
Some characters were never simplified. For example,
井 U+4E95is in TC and in SC (and for that matter, most characters). In the Unihan database, codepoints like these have neither a kSimplifiedVariant nor a kTranditionalVariant (both fields are empty).
There is a traditional version of the character, only present in TC. For example,
書 U+66F8is in TC (and not in SC). The kSimplifiedVariant of
The is a simplified version of the character, only present in SC. For example,
书 U+4E66is in SC (and not in TC). The kTraditionalVariant of
The character is used in both character sets depending on context. This is a bit of a mess, and I've written more about it here
The relations should "reflect each other", but they're not a bijection
It should be the case that if X.kTraditionalVariant includes Y, then Y.kSimplifiedVariant includes X. When I last checked the data about a year ago, there was at least one bug where that wasn't the case.
This does not mean that the relation is one-to-one or onto. There are some cases that are truly bizarre; check this out here for all of the weird cases