Your back-of-the-envelope calculations seem to be off in estimating the extent of viral load reduction in a treated versus untreated person as well as the risk of transmission for unprotected, untreated sex (5% is much higher than any other number I have seen).
Obviously the viral load is not zero because then they wouldn't be HIV positive
is not accurate. Current antiretroviral therapies are able to reduce serum HIV concentrations to be below detectable limits. That doesn't make them HIV-negative, because they still have HIV infections in their cells and their serum would test positive if they stopped antiretroviral use (i.e., they are not cured).
In 1238 couple-years, zero transmissions occurred with condomless sex (estimated 58,000 sexual interactions without a condom including both heterosexual couples and homosexual males) when the HIV-positive partner was on antiretroviral therapy to the extent that they did not have detectable serum concentrations of virus (Rodger et al, 2016). This is a sufficient observation period to make the upper limit of a 95% confidence interval =0.30/100.
You are correct that this does not mean the risk is zero, but it is very low; several people involved in that study did become HIV-positive during the study but for all of them this was through sex with additional partners not part of the antiretroviral treatment.
This study and others like it should not necessarily be interpreted as meaning that safer sex practices are unnecessary when HIV is suppressed to undetectable levels, but they speak to the efficacy of those drugs for prevention. One limitation the authors note is that because anal sex with ejaculation is associated with higher transmission rates than other types of sex, and since their study was not limited to that particular act:
despite an observed transmission rate of zero for this risk behavior, a clinically important rate of less than 2.2 per 100 couple-years of follow-up cannot be excluded. This translates into an upper limit estimate of 20% risk over 10 years.
(note this is still an upper limit estimate...further study may bring that upper limit downward; edit: there is a further follow up from this study published June 2019, Rodger et al, focusing on condomless anal sex between men; they continued to not find any cases of partner to partner transmission with antiretroviral treatment allowing them to decrease the upper CI to 0.23 per 100 couple years for the sex act known to be of highest risk for transmission without treatment)
Serodiscordant couples will need to interpret for themselves how to respond to these risks, and from a broader standpoint other issues like the consistent use of prescribed drugs are also relevant. Cohen et al, 2016 used an intention-to-treat design to study a similar question, and found several partner transmissions (fewer when treatment was started earlier), but none of these where in cases where the positive partner had undetectable serum levels, which was likely because the person chose not to take the drug for whatever reason (and still the infection rates between partners were very low than without antiretroviral treatment).
A meta analysis by LeMessurier et al, 2018 found "no transmissions with antiretroviral therapy and a viral load of less than 200 copies/mL."
Rodger, A. J., Cambiano, V., Bruun, T., Vernazza, P., Collins, S., Van Lunzen, J., ... & Asboe, D. (2016). Sexual activity without condoms and risk of HIV transmission in serodifferent couples when the HIV-positive partner is using suppressive antiretroviral therapy. Jama, 316(2), 171-181.
Cohen, M. S., Chen, Y. Q., McCauley, M., Gamble, T., Hosseinipour, M. C., Kumarasamy, N., ... & Godbole, S. V. (2016). Antiretroviral therapy for the prevention of HIV-1 transmission. New England Journal of Medicine, 375(9), 830-839.
LeMessurier, J., Traversy, G., Varsaneux, O., Weekes, M., Avey, M. T., Niragira, O., ... & Rodin, R. (2018). Risk of sexual transmission of human immunodeficiency virus with antiretroviral therapy, suppressed viral load and condom use: a systematic review. CMAJ, 190(46), E1350-E1360.
