The idea of to index or not to index (aye, THAT is the question) is that you need to know if there is a big enough benefit to outweigh the increased maintenance costs associated with the index that MUST be updated every time you add or delete a record and every time you change the indexed value for a record.
In the absence of information regarding value distribution, the cardinality (expected return percentage on a query of an indexed field) is 50% for T/F data. As tables get really big, the cost of the index gets really big and there is no really clear advantage to that index, taking overhead into account.
You, however, have information regarding distribution and you therefore know that you expect different results than the usual 50% predicted by purely statistical expectation. Therefore, this is the question you must ask:
How often will I look for the records for which those flags are FALSE?
If you will always and only look for TRUE cases, then your expected cardinality is asymmetric and skews the index/not index decision. If you expect 30K out of 400K to be TRUE then your TRUE expectation is 7.5% and your FALSE expectation is 92.5%. SO if your expectation of query "direction" is always for the TRUE case, you gain 92.5% of the time by having a query. (This is a simplified approach, not rigorously derived.)
Your other field expects 800 out of 400K which is 0.2% TRUE and 99.8% false. Again, if the queries are always in favor of seeking TRUE then you win 99.8% of the time by using an index.
The third case is the combination of the SOLD/RETURNED treated sequentially (i.e. first find SOLD then find RETURNED from that) which is 800 out of 30K which is 8/300 = 6.66% expectation of TRUE, and that gives you 93.33% improvement. Although, to be honest, SQL would probably not do them sequentially if both fields appeared in the same query.
Given that you have uneven expectations, the nature of your expected queries and searches will determine whether an index on a Boolean field is useful. Ask yourself how often you will search for TRUE and how often you search for FALSE. Then consider my comments in making your decision.
Remember, if the direction of query selection is equal between TRUE and FALSE, you lose the advantage because you in essence negated your knowledge of skewed distribution.
Hope this helps you to make your decision.