In Adult Sexual Orientation and Attraction to Underage Persons (1978), a random sample was made of 175 males convicted of sexual assault, looking at their adult sexual orientation and the sex of their victims.
The sample divided fairly evenly into two groups based on whether they were sexually fixated exclusively on children or had regressed from peer relationships.
This first group already provides evidence that half of the convicted male sexual assaulters do not come from a simple binary pool of "heterosexuals" or "homosexuals".
All regressed offenders, whether their victims were male or female children were heterosexual in the adult orientation. There were no examples of regression to child victims among peer-oriented, homosexual males.
That is, it is wrong to conclude that if a man sexually assault a boy, then he would normally be attracted to men. There is a good chance he had never been seriously attracted to adults at all, but if he he had, it was adult women he had been attracted to.
The paper suggests:
The possibility emerges that homosexuality and homosexual pedophilia may be mutually exclusive and that the adult heterosexual male constitutes a greater risk to the underage child than does the adult homosexual male.
I don't completely trust this old sample of convicted males (a subset of a subset) to exclude the chance that any homosexual men are guilty of child sexual assault. However, it becomes clear that a naive Bayesian analysis, that assumes that boys assaulted by men are being assaulted by homosexual men, is going to lead to false conclusions.
The bias doesn't stop there though. A 1996 review mentions another source of bias, which contradicts the claim that "male children are not so well protected as female children"
Current pro-active procedures to identify paedophiles detect those who victimise boys but do not detect the much greater number of paedophiles who victimise girls. Perpetrators are known to the majority of their female and male victims, and those reported are almost all male; most boys do not consider their prepubertal experiences with older women abusive.
In conclusion, it is difficult to get precise figures, due to a number of sources of biases and unknowns. (Even the binary dichotomy of homosexual versus heterosexual is clearly dubious.) However, studying the people who are caught suggests that men who are attracted to adult men are not at all in the highest risk of offending, even when the victims are boys.