Has the OP actually read the Surgeon general's report?
The report analyses hundreds of studies on the effects of second-hand smoke. What's interesting is that many of the studies individually show no definite effect (the statistics are not significant). However, it's now well recognised in medical statistics that simply reporting 'a signficiant difference' or 'no significant difference' is inadequate.
Where possible statistics should be quoted in terms of a confidence interval - which gives not only a measure of significant the risk is, but what the level of experimental error is likely to be. In the context, of the 2nd hand smoke studies the statistic is a relative risk. E.g. a relative risk of 2 means that the exposed group is 2x as likey to have the disease. For example in one study of men, with wives who smoked, the RR was 1.96 with a confidence interval of (0.7 - 5.4). Because the confidence interval spans 1 (no difference) the result is not statistically significant - so while this study does not confirm the presence of an increased lung cancer risk, it is suggestive, but is limited by large experimental uncertainty (a range of 0.7 - 5.4 is massive).
This is a recurring problem - many of the studies are based only on small population samples (e.g. they study cancer sufferes in one region, or they follow up a limited number of healthy people to see if they get any diseases).
If the data is reanalysed in a pooled fashion (e.g. data from several studies is combined and analysed together), then it is possible to reduce the uncertainty. In the case of 14 out of 16 pooled analyses, a significant result was found.
The other powerful indicator that there is a genuine effect has been the demonstration of a 'dose-response' relationship. This shows, with statistical significance, that people exposed to more secondhand smoke have a higher rate of lung cancer than those exposed to lower levels, who in turn have higher risk than those exposed to none. Again, there have been several studies showing trend with strong statistics.
The point is that to reach an overall conclusion you need to look at all the studies, their weaknesses and strong points, the pooled analyses, evidence of trends, etc. Overall, the best studies show a harmful effect, and it's on the basis of that, that the recommendations stand.