Welcome to the inequality rabbit hole. I think your intuitions are sound, and there is something odd about describing the manifestly unequal past as being more equal than the late 20th century. It's not true or false exactly, but "it depends on what you mean." This post will be mostly about conceptual measures of inequality, because that's the point on which all this hinges. The evidence, alas, cannot just speak for itself in this matter.

The fundamental problem is that we do not have a conceptually sound, robust measure of inequality that everyone agrees on. We certainly don't have one that can be easily transported across time and space for comparative purposes. The results of any attempt to do so will be strongly influenced by the methodology we choose. The choice of method, in turn, emerges from a combination of what we believe inequality to be.

The first fork in the road is whether we are interested in relative or absolute inequality. A relative measure will tell us what the income ratio is between the richest and the poorest. (A related approach would be to use shares of total income, which gives you the same idea in different language - the richest 1% earn 10% of national income.) An absolute measure of inequality asks how much more can the top income earners buy than the poorest - how many bushels of wheat/hours of labour/bars of gold can you buy with the difference between the richest person's income and the poorest? These two measures will mechanically diverge as a society gets richer. If you imagine a society with a fixed level of relative inequality, which you could describe with income shares (the richest 10% earn 50% of the national income) or a Gini coefficient (0.6, say) or any other relative measure. If that society is very poor, then the absolute gap between the richest and poorest would be small. If that society doubled its earnings, the absolute gap would increase, but the relative gap would stay the same.

To put hypothetical numbers on it, if the poorest person earns $1 a day, and the richest earns $100 a day, then the ratio is 100, and the gap between them is $99. If you double their incomes to $2 and $200, the ratio is unchanged, but the absolute income gap has doubled to $198. So, has inequality stayed the same, or has it doubled? These are radically different answers to the same question!

To return to history, this describes one of the fundamental dilemmas in comparing the Downton Abbey era with the present. In terms of absolute purchasing power, there is no question in my mind that Bill Gates is overwhelmingly richer (and has a higher income) in absolute terms than anyone alive in the 1910s. (One could use perhaps Henry Overton Wills III vs. James Dyson if you want British super-rich people. I'll stick to highly recognizable American names, the point is identical.)

Of course, goods and services have changed in price, in some cases very dramatically. If we take something whose price has increased a lot (domestic servants, land in downtown London, Dutch Old Masters) as our benchmark, then Bill Gates is not so much richer than an Andrew Carnegie or a John Rockefeller, whereas if we compare his ability to purchase things whose price has dropped dramatically (clothing, food, illumination) or which did not even meaningfully exist (computing power, television sets, mobile phones, antibiotics) then Bill Gates is so much richer it can't even really be measured. There is no single obviously correct way to reconcile this, but overall, the amount of resources Bill Gates has at his command is larger (by a lot) than even those of a Carnegie or Rockefeller. So in that sense, the absolute inequality has surely increased.

There is another perspective, one that is quite compelling based on the intuitions you've provided above, that the past sure does seem pretty unequal. Peter Lindert, Jeff Williamson and Branko Milanovic created an approach called the "extraction ratio" to create an intuitive measure of past inequality.("Pre‐Industrial Inequality" (2011) in The Economic Journal). The key thing they account for is the distance from subsistence income. A society cannot (in the long run) sustain a level of inequality such that the poorest cannot sustain themselves. Therefore, the highest level of inequality that can be sustained is bounded by how far a society is from basic subsistence. How unequal a society is should therefore be adjusted for this distance; a society very close to subsistence is closer to its maximum potential inequality at a lower level of absolute or relative inequality, and should thus be thought of as more unequal.

To give a stylized example, one can imagine a society where everyone is maximally poor (say, 1000 people earning $1 per day, which we can take to be bare bones subsistence) except for one ruler who extracts all surplus production (which amounts to $10 per day). This society will appear to be extremely equal, measured by either the relative or absolute measures of income inequality above. But in some very important sense, this society is maximally unequal; nobody except the one person earns more than subsistence. If the ruler tried to extract more revenue, people would starve, which is not sustainable in the long run. This then is the stylized case for most societies before the modern period, where measured absolute and relative inequality are both low, but yet, most people are extremely poor, and the elite take a very large share of what is available to take.

One can then imagine a society where all incomes are doubled - ordinary people earn $2 a day, and the ruler earns $20 a day. The ratio of incomes and the shares of national incomes are all unchanged, suggesting unchanged inequality. The absolute gap has increased from $9 a day to $18 a day, suggesting nearly doubled inequality. But if we compare the amount extracted by the ruler as a ratio of what could be extracted, we see that inequality has dropped dramatically. In the first case, the ruler earned 100% of the extractable income. In the second, the ruler earns only 1% of the extractable income! This is the perspective of Lindert, Milanovic and Williamson.

How does this matter for our interpretation of historical inequality? It means, in terms of potential extraction, the near past is in general less unequal than the distant past. Societies whose income grew in the early modern period, like England and the Netherlands, were becoming more unequal in absolute terms, but also much further away from the maximum inequality frontier, which is more like the second case described above. It also means that, even though the absolute and relative incomes of the very rich have increased across the 20th century, the distance of relatively poor people from the level of subsistence has also increased dramatically. As we move very far away from subsistence levels of income, the importance of this effect in understanding inequality diminishes, because the threshold of absolute subsistence becomes vanishingly small compared to our incomes. But in comparing the approximate present to 1900, and especially to earlier and poorer times, it seems to me a key point.

None of this solves the basic problem, that we don't actually have a good concept of inequality, and even if we did, we don't have great measures of it for the past or even for the present. But Lindert, Milanovic and Williamson at least provide a way to understand our intuition that the past was pretty unequal, even a Gini coefficient suggests the opposite.