I was in a similar position. For many reasons, some that are common, some that are almost unique to me alone, I was drawn to the field of oncology and really never considered anything else. However, after spending some time in medical school I was a little disappointed to find out that there is no such thing as an "oncologist." There is: radonc, heme/onc, ped-heme/onc, gyn-onc, surg-onc, and more.
I eliminated surgery and obgyn during my MS3 year. I still considered ped-onc, but realized I did not want to focus on children. I started MS4 year with a heme-onc rotation (both inpatient and then outpatient). It WAS something that I could see myself doing in the future. Much of the inpatient work- which is very rigorous - involves management of very sick patients made sicker (temporarily or not) from chemotherapy. For example, patients admitted for high dose IV chemo, patients that were neutropenic and became febrile, patients undergoing bone marrow or stem cell transplant. The management of such patients is complex and certainly requires lots of general medicine training. The selection of which drugs to give appeared to me NOT complex, and seemed rather arbitrary (based on published trials of course). I would say that if you know you want to do internal medicine first, and oncology second, heme-onc makes perfect sense, and don't give a second look to radonc. However, if your honest interest is in oncology first, look further and explore radonc, as I did, before making up your mind.
My second rotation was in Radonc. It was something I could also see myself doing in the future. It is nearly all outpatient. It is wedded to technology - both in imaging, in radiation delivery, and the merging of the two. The actual work you do in Radonc is diverse: examination and consultation with patients, careful examination of multiple imaging modalities, the physics of treatment planning, and then finally the treatment itself (which could be via a linear accelerator or sealed, implanted radioactive sources at the tumor site, or free radioactive molecules). You might also give additional, non-radiation treatments as well (for example, hormonal therapy in prostate cancer to name one). I felt that, on average, the overall fund of knowledge about cancer as a disease was greater among the physicians in the radiation oncology department than those in the heme-onc department - though there were some exceptions of course. Perhaps this is because Radoncs spend 4 years training in cancer rather than 3. Its worth noting that training in Radonc takes 5 years including internship, compared with 6 for heme-onc (3 for IM plus 3 for fellowship).
I may be going out on a limb here to my own peril, but I think the future of cancer treatment will not involve administering lots of new systemic cytotoxic drugs that put you in the hospital. As you have stated, it may involve siRNA. I would say, more generally, that it will involve increasingly targeted therapy - molecularly targeted as well as anatomically targeted. Radiation therapy is already and will be increasingly anatomically targeted as technical advances in imaging and delivery come to the clinic. Chemoembolization is another example of an (anatomically) targeted therapy. To be complete, surgery is obviously anatomically targeted, just not always practical. Kinase inhibitors and monoclonal antibodies are examples of (molecularly) targeted therapy. Perhaps we may even see cellularly targeted therapy making use of "smart" T-cells. Research in the previously mentioned treatments are fair game for anyone in any oncology specialty. I would add that either medonc or radonc residency training can be done on a research track - sanctioned by the ABIM or ABR (called the "Holman pathway"). And, at most programs that I considered, you can choose a research mentor from any department provided the research is relevant. This was true for heme-onc fellows as well.
So, the decision is up to you. Both can take you where you want to be.